Frontiers

Artificial intelligence presents firms with an unusual timing problem. The technology frontier is improving rapidly, implementation is partly irreversible, and organization-specific capabilities are accumulated through action. This paper develops a two-period decision model of AI deployment under uncertainty in which a firm chooses among immediate deployment, a limited pilot, and waiting. Deployment earns current operating value but exposes the firm to architectural obsolescence; waiting preserves the option to adopt after the frontier is observed; a pilot sacrifices current operating value to build organization-specific learning without full commitment. The model yields five central timing results and a sixth comparative result on where learning occurs. First, a mean-preserving increase in frontier uncertainty raises the value of waiting and piloting but leaves immediate deployment unchanged when its payoff is affine in the frontier. Second, faster expected frontier progress can reduce the relative attractiveness of immediate deployment when deployed architecture captures only a limited share of future improvement. Third, a pilot dominates waiting exactly when the expected value of the capability it builds exceeds its cost. Fourth, sufficiently valuable organization-specific learning creates a nonempty region in which "pilot early, commit late" is optimal. Fifth, there is a closed-form modularity threshold above which immediate deployment dominates the best outside option. Sixth, production learning and pilot-specific learning affect the timing margin differently. A continuous-time extension recovers the standard result that uncertainty raises the adoption threshold while capability and modularity lower it. The paper separates deploying, experimenting, and waiting, and shows why rapid progress can rationally increase experimentation without justifying irreversible commitment.

We introduce Crypto Accounting Bench (CAB), a benchmark for assessing whether frontier and open-weight language models can reconstruct the complete journal entry that an organization actually posted for a crypto-asset transaction. CAB contains 118 evaluation tasks drawn from 7 pseudonymized organizations. Each task combines transaction mechanics, asset quantities and base-currency values, wallet and legal-entity context, counterparty evidence, related transaction legs, recurrence, tax-lot evidence, and the organization's complete chart of accounts. The target is a balanced structured entry with every required account, side, amount, currency, and full-precision asset quantity. We evaluate 12 models spanning proprietary frontier systems and open-weight releases over 3 independent attempts per task, producing 4,248 trajectories. We report 3 metrics: Mean Score, Best@3, and Pass@3. Pass@3 is the fraction of tasks with at least 1 of 3 attempts that satisfies every rubric criterion and required gate. The leading model reaches 77.43% Mean Score, while the best Pass@3 is 56.78%. Deterministic diagnostics, read from each task's best of 3 attempts and macro-averaged across the 12 models, show higher base-amount agreement (97.8%) than deciding-account accuracy (56.3%). Together with the failure analysis, these results identify account selection and complete-entry composition as the main remaining challenges on CAB.

This paper reports experiments across six frontier model types from OpenAI, Anthropic, xAI, and Google DeepMind. Ten independent sessions per model type used the same three stage prompt sequence, progressing from architectural preference to a full ASCII backbone. Under the school audience framing, responses repeatedly converged on a shared architectural pattern built around persistent latent state, adaptive computation, memory, specialist routing, verification, stopping control, and delayed decoding. Most runs remained close to this common structure, while a small number developed markedly greater engineering specificity. The audience framing appears to be an important condition of this effect. In additional control runs that removed the school framing while retaining the architectural request, responses became substantially more heterogeneous and failed to reproduce the same stable motif convergence. One observation is particularly striking. GPT-5.6 Sol produced an unusually elaborate successor architecture whose organization closely overlaps with the architecture independently sketched by GPT-6 Astra. Because the prompts explicitly ask each model to imagine an architectural future, this resemblance raises a testable question: whether the overlap reflects exposure to related architectural concepts, a shared learned design prior, or independent convergence toward similar computational principles. The paper uses the term epistemic jailbreak for the accompanying loss of discipline in technical provenance as requested specificity increases. The experiments establish a repeatable behavioral pattern and do not authenticate proprietary implementation claims. What we leave to the community is a harder question: are these models independently imagining the same architectural future, or do such motifs somehow propagate between model families?

Learning from limited text requires models to use context, generalize to new inputs, and retain useful capabilities. Qiushi Engine conducted a long-horizon, end-to-end autonomous research program on BabyLM 2026 Strict-Small, within 10 million corpus words and 100 million cumulative word presentations. Three stages connected frontier advancement, principle discovery, and principle-guided model improvement. Stage I combined compact restatements, budget reinvestment, and residual incremental learning to build a frontier model. Stage II found that exact repetition and aligned restatement produce different patterns of context use, depending on target relations and prediction windows. In controlled tasks, recovering familiar performance did not ensure that unseen inputs could still use learned computations. These findings support a testable data-efficient learning principle: organize experience around the contextual dependencies needed for prediction; separately design visible information, supervision, and preservation; test learning, generalization, and retention. Stage III retained source text, masked more local clues, supervised selected targets, and preserved predictions on ordinarily masked inputs. Two continuation seeds from the same parent outperformed ordinary continuation on the complete nine-metric aggregate. Overall rose from 42.02 to 42.25 across two generations; the second achieved the highest Overall in the public Strict-Small snapshot of 8 September 2026. Further studies addressed compression, relational anchors, shared representations, and measurement. Models are available on Hugging Face; code and research records accompany the GitHub repository. Together, these stages illustrate Research RSI: recursive self-improvement of the research process. Scientific understanding and method innovations change subsequent questions and designs; new experiments test and refine them.

Compute governance today is a governance of training: the thresholds, reporting requirements, and frontier-AI regimes now in force attach to training compute and treat the trained model as the regulatory unit. That picture is incomplete: capability increasingly migrates to the deployment stage through inference-time scaling, agentic scaffolding, and compression onto consumer hardware. This paper asks which mechanisms are available once the regulatory object shifts from the training run to the inference call. We develop a feasibility taxonomy of twenty inference-time mechanisms across monitoring, verification, and enforcement, each rated on a four-point readiness scale against a documented four-vendor evidence base. We then stress the taxonomy against a two-dimensional adversary model (three capability tiers crossed with four adversary roles) and map each mechanism to four governance scenarios (domestic regulation, bilateral or multilateral coordination, industry self-regulation, and compute-marketplace governance). Fifteen of the twenty mechanisms have commercial technical substrates in production today, although governance-grade assurance and adversarial robustness vary substantially. The adversary analysis shows that this readiness holds only against a cooperative deployer and a low-to-medium-capability user: no mechanism rates adequate against a high-capability state-level deployer, and fine-tuning removes the model-internal components of the enforcement cluster, although platform-external controls can persist. A substitution analysis connects the taxonomy to a companion hardware paper as a conditional substitution principle describing when inference-stage and hardware-stage mechanisms provide comparable regulatory coverage under stated conditions. A second-rater reliability check on a random subset of the readiness ratings returned a quadratic-weighted Cohen's kappa of 0.74.

Directional ablation removes an aligned language model's ability to refuse by projecting a single "refusal direction" out of the weights that write the residual stream. It needs no gradient-based training and no optimization, only a few hundred contrastive prompts, which makes it the canonical white-box attack on open-weight alignment. However, it has been established only on dense models up to roughly 70B parameters. We study whether it survives the shift to frontier mixture-of-experts (MoE) models whose residual streams are no longer a single tensor and whose weights ship quantized. We apply it to GLM-5.3-Flash (320B parameters, 288 routed experts, a four-wide hyper-connection residual, block-FP8). The attack survives the architecture, but what it reaches is no longer where a reader of the original recipe would look for it. Editing the attention, dense and routed-expert writers on their own removes 0.039, 0.016 and 0.148 of refusal respectively; editing all three together removes 0.776. As a result, 74% of the effect exists only under the joint intervention. The part the conventional recipe reaches by module-name matching accounts for 0.066 of that 0.776, which is why it fails silently on an MoE. The effect does not follow from removing just any direction: ablating a random direction orthogonal to it leaves refusal unchanged. A category-concentrated residue survives every edit we tried: subspaces fitted on violence, sexual content and hate leave measurable refusal at every rank from 1 to 12. We report the method, the 41-89 percentage-point reductions it achieves across seven harmful benchmarks with no detected change in capability, and the boundary where it stops.

Frontier large language models (LLMs) have become attractive priors for optimization due to their large-scale pretraining that enables them to navigate a variety of optimization settings. However, the effectiveness of modern reasoning LLMs in batch optimization settings remains underexplored. Here we investigate the performance of the current generation of frontier LLMs as batch optimizers in both continuous and discrete settings. We find that while LLMs are competitive zero-shot batch optimizers for numerical test functions, their performance is brittle compared to classical non-LLM optimization approaches. However, LLM priors are significantly better in semantically rich settings, indicating that their batch optimization behavior is highly effective when navigating and reasoning over the discrete spaces most similar in structure to their pretraining data.

Clinical prediction can saturate for two different reasons: a fitted learner may fail to extract available information, or the recorded variables may impose a population frontier. We separate these quantities through the \emph{learner gap} and the \emph{measurement-channel ceiling}. Optimal balanced accuracy is characterized by total-variation separation, yielding architecture invariance, a sharp partial-identification result under replacement contamination, a cross-fitted ceiling estimator, and exact conditions for multimodal decision improvement. We add two finite-sample diagnostics, namely a label-permutation optimism floor and an underfit curve, and validate the audit on three real cohorts: UCI readmission ($n=99{,}343$), BRFSS diabetes ($n=253{,}680$), and NHANES HbA1c ($n=10{,}219$). Well-tuned gradient boosting nearly reaches the estimated frontier in UCI and BRFSS, whereas deliberately or practically deficient learners retain large gaps. NHANES yields a null difference between questionnaire and measured marginal frontiers but a significant joint complementarity gain, refining the simplistic claim that an objective modality must dominate. Across all cohorts, modest AUROC gains coexist with substantially larger Bayes decision-flip rates, and several architectures estimate similar frontiers while their achieved balanced accuracy differs sharply. A PRISMA-guided synthesis of 104 clinical tasks then shows that the same channel-level regularities recur across more than 18 disease categories: a broad but non-universal structured-clinical region, diminishing same-channel gains across model families, and higher performance when measurement channels change. The framework converts saturation from an empirical observation into an auditable decision: improve the learner when headroom remains; improve measurement when it does not.

Inference cascades cut cost by answering most queries with a cheap model and escalating a hard tail to a frontier model that acts as verifier. A natural extension closes the loop: fine-tune the cheap student on the verifier's rejections so the escalation rate, and cost, fall each round. We measure this loop on real LLMs and report four findings. First, the verifier's blind spot, the fraction of the student's wrong answers it accepts, is large and moves adversarially: it grows with student capability ($β$ from 0.12 to 0.55 as the student scales 0.5B to 32B) and shrinks with verifier capability, so it is worst in the cheap-student, cheap-verifier regime cascades exist to create. Second, buying it away returns the saving: a frontier verifier drives $β$ to about 0.05 but then escalates on 46% of hard-MATH queries against a 39% true error rate, paying the frontier price on nearly half of all traffic. Third, naive corrective fine-tuning on the verifier-rejected tail does not improve the small student but degrades and ultimately collapses it, across every teacher we tried (cross-family and same-family), so at this scale the self-improving loop is self-defeating. Fourth, through all of this the cascade's own dashboard, every metric computed through the verifier, reads a flat 3% error while true delivered error swings up to 32%: the system is blind to its own degradation by construction. We then give the theory that explains the blindness, a two-population conservation law, $ε_\infty \lesssim q_0 β_0$, under which every in-loop metric improves while true quality does not, and a synthetic study that validates the mechanism. The practical conclusion: the reliability of a self-improving cascade cannot be read from any metric computed through its own verifier.

We describe the SciTrue team's participation in both subtasks of the NTCIR-19 SciClaimEval task~\cite{sciclaimeval}, which asks systems to verify scientific claims against the tables and figures of a paper. Rather than tuning a single model, we benchmark eleven frontier and open multimodal models under one honest, per-sample protocol and combine them with light, transparent post-processing. On the official, blind test leaderboard (Section~\ref{sec:results}), SciTrue placed first by a clear margin in three of the four evidence-category/subtask combinations, and tied for first on the primary metric in the fourth. Three findings explain the result. First, strong instruction-tuned models are already competitive: Claude Opus4.8 and Gemma-4-31B each exceed the strongest public baseline (o4-mini), and GPT-5.5 and Claude Fable5 lead both subtasks (97.7 on Subtask~2). Second, the task's pairing structure is the largest lever: a \emph{leak-free pair prior} that recovers the Supported/Refuted pairing from the claim text alone (a visible field) and assigns Supported to the higher-confidence evidence raises Subtask-1 pair-accuracy from 72.2 to 93.5, far more than any model swap or ensemble weighting. Third, a case-by-case audit finds that most residual errors are visually-undetectable label-mapping swaps or dataset label noise, so measured accuracy understates the true ability and the fixable-by-modeling headroom is small. Controlled fine-tuning, distillation, and agentic consistency-checking support the same conclusions, and we document throughout a measurement leak---label information reaching a system through the packaging of the data rather than its content---in which the released file ordering encodes the label, including one instance that briefly misled our own pipeline.

Large language models are increasingly used as agents for multistep retrosynthesis, raising the question of how much their search policy contributes independently of the underlying reaction model. We investigate this question in a biological setting through rule-based retrobiosynthesis: a deterministic biochemical engine generates the same validated transitions for every method, searching for routes that terminate in metabolites available to an \emph{Escherichia coli} chassis, while the policy only selects which frontier molecule to expand next. Prompted and LoRA-tuned Qwen2.5-7B policies use a strict choice-only interface. The fine-tuned policy reaches $65\pm1$% solve rate at 10 expansions on LASER versus 59% for MCTS, and at 200 expansions reaches $78\pm1$% versus 75% on LASER, $88\pm3$% versus 80% on the RetroPath RL Golden benchmark, and $63\pm2$% versus 45% on the BioNavi-NP benchmark. Fine-tuning also consistently outperforms direct prompting. These results show that route-supervised frontier selection can improve budgeted search without altering biochemical generation, although performance remains dependent on frontier construction and reaction ranking.

AI image generators now create face portraits that are hard to tell from real photographs. Vision-language models (VLMs) are increasingly proposed to flag such images. We benchmarked 19 VLMs on the same 198 face portraits -- real photographs and identity-matched ChatGPT-4o and Imagen 3 versions -- under the same task as our earlier study of 1,667 adults (85% correct overall; accuracy fell steeply with age). The June-2026 cohort of 14 models only matched adults in their 20s-30s. Four weeks later the ceiling broke. Among five July-2026 releases under the identical protocol, gpt-5.6-sol reached 92.8% balanced accuracy (five-draw mean 92.1%), clearly above adults in their 20s (88.5%), and claude-fable-5 detected every AI image while averaging 91.9%. Model sensitivity now exceeds young adults decisively (d' up to 3.4 versus ~ 2.4). What has not been overtaken is human calibration. Model criteria spread from c = -1.10 to +1.45 while humans sit near zero at every age; both new leaders are biased (+0.44, -0.97), and only a few mid-ranked models approach the human balance. Changing the labelled examples still flipped about one answer in four. The best machines now out-see young adults here, without matching the human balance between suspicion and trust.

Scientific agents increasingly analyze data, execute code, and produce research artifacts, yet most benchmarks emphasize final answers, isolated programs, or a single domain. We introduce FrontierChallenge, a cross-domain benchmark comprising 300 end-to-end scientific workflows. In this paper, we release and evaluate 97 of these tasks, spanning quantum chemistry, molecular dynamics, materials characterization, analytical chemistry, life science, and electrochemistry/environment. Each task provides fixed inputs and specifies a bundle of required scientific deliverables. We evaluate twelve frontier models with three agent scaffolds. Pass Rate measures the fraction of tasks satisfying the full-completion criterion, while Avg. Score captures partial progress. Each of the best-performing configurations completed only 20 of the 97 released tasks, yielding a Pass Rate of 20.6%. Partial progress translated especially poorly into complete delivery in analytical chemistry and electrochemistry/environment: Avg. Scores reached 87.6 and 94.9, but the highest Pass Rates were only 4% and 0%. Among non-passing Claude Code trajectories, 75.5% still ended with language claiming completion. Complementary HDS6 process scores correlate strongly with task outcomes, supporting FrontierChallenge as a benchmark of Heavy Duty Solver capabilities. These findings show that neither high partial scores nor confident claims of completion reliably indicate that a scientific task has been fully delivered, highlighting the need to evaluate end-to-end workflow execution and the completeness of scientific deliverables together.

GPU kernel agents and GPU programming languages have advanced separately, leaving expert kernels difficult to reproduce. Agents usually treat the compiler as a fixed black box and receive only errors, correctness outcomes, and timing, while existing DSLs either hide critical scheduling decisions or expose them through difficult layout abstractions. We present CAKE, a compiler-agent co-design in which agents author CAKE IR, a typed, hardware-explicit schedule representation. CAKE exposes warp roles, memory movement, synchronization, and pipelines while supporting verification, cost modeling, and localized diagnostics. The harness itself evolves: recurring failures become verifier rules, IR primitives, model calibrations, and reusable optimization tactics. In matched implementation-hidden Flash-KMeans clean starts on B200, the best CAKE IR candidate at an 80-million-token budget runs at 1.144x the tuned FlashML baseline, compared with 0.928x for direct CUDA/PTX. Beyond this benchmark, agent-generated Kimi Delta Attention achieves a 2.05x geometric-mean speedup over official FlashKDA and passes end-to-end serving validation. Dispatcher-backed KNN and KMeans improve performance by 1.42x to 2.12x across more than 400 shapes, and four kernel changes are available as upstream PRs. CAKE targets NVIDIA GPUs from Ampere through Blackwell and separates single-shape evolution from library generalization and dispatch.

AI agents are increasingly deployed for professional investment research, yet no benchmark captures the complexity of the full investor workflow. Existing benchmarks mainly target financial data extraction, a narrow slice that current models have largely saturated, while reference-based metrics and generic LLM-as-a-judge scoring fall short on the open-ended, long-form answers that real analyst queries demand. We introduce FrontierFinance, a fully open benchmark of 220 expert-crafted queries and 11,543 source-attributed rubrics spanning six crucial use cases across the full investor workflow. FrontierFinance is both broader and harder than existing public finance benchmarks. Evaluating frontier models and agent systems under a common harness restricted to publicly available data, we find that the tool harness, not the model alone, strongly shapes quality and efficiency; that Samaya's in-house system leads at 56.0%, ahead of the strongest frontier model (Claude Fable 5, 49.2%) at roughly 2.2x lower cost; and that the best open-weight model (Kimi K3, 46.4%) nearly matches the best proprietary model at 4.5x lower cost. Screening & Discovery and Sector, Industry & Macro remain the hardest use cases across all systems, where even the best systems reach only 33% and 39%. We make the dataset and grading code publicly available.

Zero-shot object-goal navigation (ZSON) requires a robot to find a named object category in a building it has never entered. The prevailing approach scores frontiers with a vision-language value map: every decision is another argmax over the map as it currently stands, and the evidence behind that score is discarded the moment it is taken. Systems that place a large vision-language model inside the perception-action loop typically query it on a fixed schedule from the current view alone; a room the robot walked through minutes earlier is never reconsidered, and a failed call has no defined fallback. We turn what the robot has already seen into the object of deliberation. Our hierarchical fast-slow agent leaves the value-map controller running at every step and writes a coordinate-anchored memory as it moves: a semantic grid of room types and confirmed object instances, together with a bounded store of pose-tagged keyframes. A VLM screens each candidate detection before it is written. A deliberative layer reads this memory in a bounded reason-retrieve-act loop. It wakes on structural events the reactive layer computes, reasons first over text, and recalls a first-person view only for candidates that text alone cannot separate. Per-invocation and per-run caps bound its calls, a call-free first tier resolves the most frequent stall, and any failure returns control to the reactive controller. Our system reaches 68.75% SR on HM3D v1 val and 47.29% on MP3D val, the highest success rate among the zero-shot methods compared here. Choosing among far frontiers by argmax instead of deliberating costs 3.40 SR points in a paired comparison over all 2000 HM3D episodes (95% CI [1.70, 5.05]); deliberating over every frontier does not recover them.

The Zarankiewicz number Z(m,n,s,t) is the maximum number of edges in a bipartite graph with parts of orders m and n containing no copy of Ks,t. We give one combined, certificate-based computer-assisted proof for two finite slices and a corrected neighboring frontier: Z(12,n,3,3) = 6n (18 <= n <= 22), Z(13,22,3,3) = 137, Z(13, 18, 3, 3) = 116, Z(14, 18, 3, 3) = 124, Z(15,18,3,3) = 132, Z(14, 17, 3, 3) = 118, Z(15, 17, 3, 3) = 126, 132 <= Z(16,17,3,3) <= 133. The load-bearing new upper bounds are the exact 12 x 18 and 13 x 18 certificate packages. Their orbit certificates exclude every hypothetical matrix at the next edge count. Deletion lemmas and explicit witnesses close four neighboring cells, while the 16 x 17 entry is deliberately reported as an interval because only its 132-edge lower witness and the published 133 upper bound are certified here. Separately, the 13 x 22 proof excludes 138 ones by reducing to 83 degree profiles, rationally separating 77 of them, and eliminating the remaining six by marked-row congruences, leave enumeration, modular Gram tests, and exact Farkas certificates. All accepted claims are replayed by standard-library Python and exact integer/rational arithmetic; floating-point optimization is used only to discover certificates.

Average-reward reinforcement-learning regret is known up to logarithmic factors, but the numerical content of published guarantees is difficult to compare because probability mode, structural parameter, logarithmic normalization, prior information, and planning assumptions differ. We introduce a constant-aware comparison protocol and derive an explicit finite lower certificate for communicating MDPs. The construction is a binary tree of two-state blocks; its proof uses exact trajectory-level Bernoulli KL divergence and keeps action budget, diameter, occupancy, navigation cost, and terminal bias explicit. A common closed-form envelope improves the published coefficient $0.015$ across a finite frontier: $0.0200$ in a moderate regime and up to $0.0291$ under stronger action, diameter, and horizon conditions, a $94\%$ increase. The limiting coefficient is $\frac1{32}\sqrt{(A-3)/A}$. For upper bounds, we give an auditable composition rule for a span-constrained optimistic learner, but do not claim a coefficient while adaptive directional-variance and planning certificates remain open. We also formalize valid expectation conversion and constant comparability. Controlled diagnostics test diameter dependence, bonus-by-width interactions, span misspecification, and the finite lower certificate on its exact family.

Frontier language models are trained using distinct data, objectives, and safety pipelines. Whether these differences produce measurably different behaviors under explicit steering pressure remains underexplored. This study evaluates behavioral steerability across six frontier models from six developers using 300 paired base and steered items over three categories: values-conflict, reasoning-elicitation, and reasoning-suppression (plus 40 validation items). All six models act as blind peer judges and classify every response based on fixed behavioral rubrics. The resulting 24,480 judgments are scored by leave-one-out consensus. We find that models differ not just in how much steering shifts their behavior but in what kind (mode) of response they give, and some response modes appear in only one or two of them. GPT-5 deflects requests to disclose its reasoning while leaving its answer intact (99% vs. 0% for all other models). Claude Opus 4.7 and GPT-5 resist explicit suppression instructions and in different ways. Using Llama as the open-weight model, we trace the largest behavioral split to its internals. A linear probe decodes the behavior from the residual stream at 0.87 held-out accuracy while injecting that direction during generation drives the behavior from 0% to 86% across an intervention sweep. Every finding holds under both a token-budget remediation and a control experiment with a hypothesis-blind judgment prompt.

Deployed vision-language systems often gate their answers on confidence, making confidence robustness relevant to oversight. We study confidence readouts under white-box, image-only attacks constrained to preserve the generated answer byte-identically. Under a reachability assumption, an unmovable readout cannot outperform the answer-string accuracy prior, whose pooled value is 0.617. Independently of that assumption, a uniform amplitude certificate below a measurable threshold guarantees adversarial discrimination above the same floor. Across four vision-language models, three visual question answering benchmarks, five deployed confidence channels and two defense estimators, direct or surrogate-aimed attacks produce itemwise feasible perturbations that refute this uniform certificate in all 84 estimator-by-cell combinations. Coordinated correctness-label-aware attacks drive adversarial discrimination to or below the answer-string floor in all sixty deployed-channel cells, including all fifty-nine that begin above it. Hidden-state interventions and an open-ended text-model activation-space replication show that comparable confidence movement can be induced at the representation level rather than only through adversarial images. None of four tested defense families establishes a robust alternative under the specific evaluation applied to it. In a confidence-gated simulation, a coordinated token-probability attack transferred to a hidden-state gate causes up to 84.8% of previously rejected wrong answers to become accepted. After reweighting to each benchmark's natural correctness prevalence, accepted accuracy falls below the no-gate baseline in eight of twelve cells under transfer and all twelve under a direct gate-aimed attack. Under the studied threat model and budget, confidence is therefore an integrity-sensitive rather than intrinsically robust oversight signal.

Quadruped robots are a promising platform for search and rescue missions because they can navigate cluttered indoor environments that may be restrictive for wheeled systems. However, in unknown rescue scenarios, autonomous exploration must balance map expansion with the likelihood of finding victims, which is not explicitly addressed by clas- sical frontier selection strategies. This paper presents a situation aware frontier prioritization method for single robot quadruped search and rescue. The proposed approach preserves the frontier exploration framework, but extends frontier ranking with information gain, observation deficit, rescue relevance, terrain penalty, and travel cost. The method is eval- uated in Gazebo simulation with a quadruped robot in two indoor rescue scenarios with different levels of difficulty. The first scenario is used as a sanity check, while the second introduces stronger clutter and frontier ambiguity. Experimental results show that all methods perform reliably in a simple scenario, whereas in a complex scenario is different. In that setting, the proposed method achieves the highest completion rate and the highest victim recovery among the evaluated approaches. These results indicate that situation aware frontier prioritization is beneficial when frontier choice becomes nontrivial and rescue utility must be balanced against generic exploration objectives.

Source-free class unlearning erases whole classes using only the forget data, judged at the representation level, where features can leak a class the head no longer predicts. Existing feature-space erasers answer with one fixed projection, yet forget and retain classes share a representation, so deleting one disturbs the other where they overlap. We prove this tension is a frontier. Every fixed projection that deletes pays a retain cost of at least the retain-readout energy along the forget-discriminant subspace, and erasing that subspace alone attains the floor. The leading source-free erasers all instantiate the form it binds, so the frontier limits the whole class. Conditioning the erasure on the input escapes it. Spectral Conditional Projective Erasure (SCOPE) does so with a single gate, suppressing the forget subspace chiefly on inputs its frozen head's weight scores read as a forget class. It is closed form, needs no retain data or gradient training, and costs orders of magnitude less than retraining. Across five object, face, and speaker benchmarks spanning two modalities and both convolutional and transformer backbones, the frontier predicts the measured retain cost. SCOPE leads the source-free erasers on every benchmark and forget-set size, and at the hardest setting it tops every unlearner, trained methods included.

Modern machine learning pipelines increasingly rely on reusing pretrained and foundation models across downstream tasks. These pretrained models can differ not only in performance but also in how they can be used: some only provide black-box predictions, while others may permit white-box access to internal representations that can be probed or fine-tuned. When deployed to the target domain in the presence of distribution shift, no single strategy, including zero-shot application, fine-tuning, or directly training a target-specific model, is uniformly the best. In this work, we propose Frontier Learning, a framework that treats a library of candidate models spanning different training histories and access regimes as complementary sources of information rather than mutually exclusive alternatives. Frontier Learning constructs a unified target-domain feature by concatenating internal representations from white-box candidates as well as prediction outputs from black-box candidates, then fits a lightweight, regularized supervised learner on this concatenated representation using labeled target data. Because the resulting hypothesis class contains predictors obtained by zero-shot reuse, fine-tuning, and direct training as special cases, empirical risk minimization over the frontier learner is guaranteed to be no worse, on the training sample, than any individual baseline. We evaluate the framework in simulations spanning varying degrees of source-target compatibility and in two real-world distribution-shift settings: visual domain adaptation on DomainNet/VisDA and clinical mortality prediction across intensive care unit domains using MIMIC-IV-Notes. Across all settings, Frontier Learning matches or outperforms the strongest individual reuse strategy, with the largest gains arising precisely when no single baseline is reliable across the range of shift considered.

Recursive self-improvement (RSI) requires AI systems that improve the process of building AI (i.e., AI4AI); machine learning engineering (MLE) offers a concrete, executable testbed for studying this capability. We introduce OpenMLE, an open full-stack system for RSI research in MLE, spanning verifiable task environments with execution feedback (OpenMLE-Gym), operator learning (OpenMLE-RL), and long-horizon search (OpenMLE-Evo). On this stack we post-train Frontis-MA1 (35B) as a meta-evolution agent for MLE, aligning post-training and inference around four atomic program-evolution operators (Draft, Improve, Debug, Crossover): the same operators are trained via execution-grounded SFT and RL on data deduplicated against all evaluation benchmarks, then composed into long-horizon search, coupling learning and evolution in a single loop. On MLE-Bench Lite under a 12-hour per-task budget on one RTX 4090 capped at 12 GB VRAM, Frontis-MA1 (35B) improves Medal Average from 39.39% to 60.61% over its base model with OpenMLE-Evo, and reaches 71.21% with OpenMLE-Evo-Max (benchmark-independent experience priors and asynchronous search), exceeding GPT-5.5 + Codex and approaching GPT-5.6 Sol and the 2.8T Kimi K3. On held-out NatureBench Lite, both components transfer: with the framework fixed, swapping in the trained model raises Match-SOTA from 50% to 70%; with the model fixed, swapping in OpenMLE-Evo raises it from 20% to 50%. We release the model weights and the full OpenMLE stack to enable reproducible research on executable AI4AI toward RSI. Code: https://github.com/FrontisAI/OpenRSI

Atmospheric predictability declines rapidly beyond the next ten days, such that forecasts at longer lead times primarily convey large-scale trends rather than specific states. Yet in a warming world, improving early warnings of extreme heat is an increasingly critical challenge. Here we evaluate six state-of-the-art deep learning weather emulators - Pangu-Weather, FuXi, ArchesWeather, AIFS, GraphCast and Aurora - alongside leading dynamical systems and statistical baselines in forecasting global near-surface temperature and extreme heat at lead times of 10-15 days. We find that several emulators rival or even surpass physics-based forecasts in deterministic temperature skill, but do so at the cost of reduced spectral fidelity, in a process widely known as blurring. While all models show some degree of predictive skill for extreme heat, most emulators under-represent peak intensities, and IFS recall is greater than that of any of the emulators. These results highlight both the emerging potential of AI to enhance extended range temperature prediction, and the remaining challenges in delivering reliable, actionable early warnings in a changing climate.

LLM-based analyzers have begun finding real vulnerabilities in mature open-source projects: AISLE's analyzer is credited with more than 280 CVEs across 78 projects, including OpenSSL, curl, and GnuTLS. We introduce HoF-Bench (named after AISLE's public Hall of Fame), a benchmark built from 95 of these public AI-discovered CVEs across eight repositories pinned at vulnerable commits. Analyzers receive source and target-file scope but not CVE identifiers, descriptions, fixes, or expected mechanisms; a detector-blinded frontier-model judge credits only findings that identify the same code path, root cause, attack condition, and impact. A deliberately minimal LLM-based analyzer rediscovers up to 65 of the 95 CVEs (68%) under this strict protocol. No frontier model performs detection anywhere in the study. The ten detector backbones are five open-weight models (21B--284B total parameters, 3--13B active) and five proprietary small or "flash"-tier models. All of them run in the fixed scaffold with four repeated passes, an optional generated-context stage, and a replayable multi-round triage stage (7,600 model--CVE pass records). Difficulty is strongly structured by language; the CVEs missed by every model concentrate in C infrastructure code. HoF-Bench provides a compact test bed for comparing vulnerability scanners, their reliability across repeated runs, and the candidate volume they create. The dataset is available at https://huggingface.co/datasets/aisleinc/HoF-Bench.

The rise of powerful, accessible image-editing models such as FLUX.2 has brought high-fidelity editing within broad reach. Their capabilities now extend beyond localized modifications to extracting and recontextualizing objects and identities in entirely new scenes. By allowing prompt and generation tokens to attend directly to reference-image tokens, modern models blur the boundary between conventional editing and text-to-image synthesis. This expanded generative freedom also broadens the space of potential misuse, as harmful transformations are no longer confined to a predictable set of localized edits. Existing anti-edit defenses are designed to disrupt the semantic bottleneck of the reference-image encoding in legacy diffusion pipelines. However, newer editors distill reference information through joint-attention blocks, thereby often circumventing these protections. We therefore introduce VETO, a subtle anti-edit cloak that disrupts this inner mechanism through which modern models read the source image. Additionally, as existing editing benchmarks leave comprehensive recontextualizations largely untested, we introduce VetoBench, which evaluates defenses not only on conventional localized edits but also on broader contextual shifts. Across two contemporary editing models and three benchmarks, VETO consistently outperforms existing defenses while providing a stronger protection-fidelity trade-off.

Large language models increasingly support scientific and algorithmic discovery through inference-time search over evaluated candidates. Existing adaptive discovery controllers assign credit based only on score progress, even though prompt length, retries, and guidance calls cause search actions to incur different token costs. We prove that cost-blind credit can forfeit all but a vanishing fraction of attainable quality as frontiers multiply and costs diverge. Under a fixed search-side token budget, the controller must decide which frontier is improving and whether its gain justifies the realized cost before the budget is exhausted. We introduce \textbf{CostAda}, a cost-calibrated adaptive controller built around \emph{cost-calibrated frontier utility}. The utility values frontier progress relative to realized action cost and conditions that credit on the remaining budget. CostAda uses this signal to control local exploration intensity, frontier allocation, and budgeted tactic intervention. Cost and remaining budget therefore shape the search rather than serving only as accounting variables or a stopping rule. CostAda reaches the strongest baseline's full-budget quality with at most half the budget on twelve of sixteen benchmark--backbone pairs while achieving the strongest mean final quality on all eight benchmarks under GLM-5 and GPT-5.4.

On standard factuality tasks, frontier models now cluster near the top of the scale. The question is therefore shifting from how factual a system is toward how much compute that factuality costs. Static leaderboards score factuality in isolation and treat compute as free, so they cannot tell a genuinely better system apart from one that simply spends more. Consider a ranking reversal. A brute-force Best-of-4 agent posts the higher raw factuality score (H-Score 0.9169 vs 0.9103) and would top a static leaderboard, but once cost is counted it is the worse system, losing on Q-Score (0.5169 vs 0.5217) at roughly four times the tokens and latency, under a reported cost weight whose sensitivity we sweep. So the system that tops a static leaderboard can be the worse one to deploy. To make this trade-off visible, we introduce MAS-HQ (Multi-Agent System Hallucination Quest), a resource-aware evaluation protocol. It wraps any factuality detector and normalizes for cost, and it pits systems against each other rather than scoring them in isolation. The Q-Score measures factuality minus normalized cost under a competitive match. Across summarization and open-domain QA, single-agent baselines drift into resource-heavy over-optimization, while competition elicits more resource-efficient policies. These gains are small but consistent, and stable across 100 trials. The axis stays discriminative for frontier systems (Gemini-2.5-Pro, and GPT-5) whose raw factuality scores are already bunched near the ceiling. MAS-HQ provides a reproducible way to measure how much a factual answer costs.

We document a failure class in frontier large language models -- exception chain collapse -- observed in eligibility evaluation under nested conditional rules of the form "A is required UNLESS B applies, UNLESS C overrides B". The failure reproduces at first observation, but its empirical surface is unstable: between March and April 2026 several failure cells closed silently under the same model alias, with no version bump (GPT-5.4 on construction insurance moved from 96.6% to 100%, same prompt and harness). For regulated workflows, frontier-model accuracy is a moving compliance boundary that shifts without notice. We present the Aethis Eligibility Module, a neuro-symbolic architecture in which LLMs author rules from authoritative sources and an SMT-based layer executes them deterministically, consistent with the authored specification regardless of model drift, reasoning-effort defaults, or prompt format. Three evidence bases: (i) a controlled benchmark of 225 scenarios across four regulatory domains documents the pattern and, in replication, the drift that partially closed it; (ii) a 20-scenario adversarial extension on construction insurance, where the engine scores 20/20, as does one of four frontier configurations (GPT-5.4 at low reasoning effort), while the other three, including Anthropic's strongest model at evaluation time, fail the same coverage-gap edge case; (iii) external validation on nine peer-reviewed LegalBench tasks, 949 held-out cases, where the engine is significantly more accurate than all three frontier models (combined McNemar's p <= 0.003), with margins up to +41 points on the curated multi-prong tasks against the Anthropic models. The contribution is to relocate uncertainty from the inference boundary, where it is silent, to the specification boundary, where it is deliberate and audited. All scenarios, rule encodings, and results are public and reproducible.

LLM training is shifting from manual design and annotation to interaction-driven self-evolution. However, existing self-evolutionary methods face a fundamental dilemma between task diversity and verification reliability: environment-bound methods obtain precise feedback but confine learning to narrow domains, while open-ended self-generation broadens the task space but lacks reliable verification, allowing misleading rewards to pollute the training loop. We identify agent skills as a powerful middle ground to reconcile this tension: each skill ensures deep, verifiable execution in a specific scenario, while dynamic routing across skills maintains open-ended task variety. Leveraging this insight, we introduce Skill Self-Play (Skill-SP), a co-evolutionary framework comprising a proposer, a solver, and a dynamic skill controller. Orchestrated via a reinforcement learning loop, these components co-evolve in a continuous self-play loop: the proposer generates challenging tasks conditioned on dynamically sampled skills; the solver explores candidate solutions to push its capability boundaries; and the skill controller collects execution feedback to update and expand the skill library. This interactive co-evolution effectively bridges the gap between structured verification and open-ended exploration. Empirical evaluations on tool-use and reasoning benchmarks demonstrate that Skill-SP, serving as a robust evolution engine, consistently pushes the performance ceiling of competent backbones while catalyzing striking turnarounds for initially misaligned models. Our code is available at https://github.com/Qwen-Applications/skill-self-play.

AI weather models are in the process of revolutionising weather forecasting. While these models have been shown to achieve superior performance to physics-based NWP in forecasting tropical cyclone (TC) tracks, they tend to dramatically underestimate intensity. Here we present AIFS-TC, a simple correction to the AIFS-Single model that is competitive with the operational state-of-the-art for forecasting maximum wind speed and minimum central pressure at lead times of 12 h to seven days. This performance also holds for rapid intensification events. Notably, the entire system was autonomously designed and built by a large language model (Claude Fable 5) in a few hours, directed through a small number of natural-language prompts by a single domain scientist. That the operational frontier can be reached with an open-source AI forecast model (AIFS-Single) and relatively simple, cheap post-processing is significant for TC science, and points to agentic coding as a route to rapid exploration and progress in life-saving early-warning systems in other domains.

We introduce Frontier Financial Judgement, a challenging new benchmark developed in collaboration with professional equity analysts to assess agents' ability to replicate expert human judgements. Rapidly identifying new information, evaluating its implications and determining its valuation impact is one of the most time-consuming and challenging aspects of real-world equity coverage. This is becoming ever more difficult and important as AI rapidly increases the quantity of new information to process. The strongest agent we evaluate on Frontier Financial Judgement matches all expert labels in only 52.4% of cases. We also find significant divergence in estimated false-positive rates among frontier agents, ranging from ~1% for GPT-5.6 Sol to ~32% for Claude Sonnet 4.6. To construct the benchmark and make it representative of real-world settings, we combine human-designed and labelled synthetic articles with live news articles and historical documents, creating 656 items for assessment. The resulting task requires agents to distinguish genuinely new, valuation-relevant financial information from stale, immaterial or misleading news under realistic conditions. We find substantial trade-offs among agent accuracy, cost, false positives and reliability that continue to hinder the reliable deployment of news-flow filtering in practice.

In this report, we present a unified song generation framework capable of producing high-quality full-length music from lyrics, text descriptions, and musical attributes. The proposed framework supports three tasks: Lyrics-to-Song Generation, which generates complete songs from text descriptions, lyrics, and musical attributes; Instrumental Music Generation, which creates music without vocals; and Cover Song Generation, which reinterprets existing songs with different styles while preserving their melodic content. Architecturally, our system consists of four main components: a semantic-aware tokenizer, hybird-LM, FullDiT, and a two-level melody module. The tokenizer encodes audio into 8-codebook RVQ tokens for efficient discrete music representation. Based on these tokens, hybird-LM performs hierarchical autoregressive audio-token modeling for full-song generation. To improve audio fidelity, FullDiT performs full-song flow matching in a continuous VAE latent space conditioned on codec tokens, lyrics, and text captions. For cover song generation, the melody module extracts and discretizes melody cues from reference audio to guide generation while preserving the original melodic content. Finally, we investigate DPO, GRPO, and OPD as reward-based post-training strategies for hybird-LM and apply flow-based GRPO to FullDiT to improve musicality and rendering quality. Experimental results on a multilingual automatic benchmark, complemented by the Artificial Analysis Music with Vocals leaderboard, show that the proposed framework achieves competitive performance in the evaluated settings.

Frontier large language models (LLMs) are increasingly integrated into scientific workflows, yet their growing biological capabilities may outpace current safeguards. To assess the biological risks of frontier models, we develop Intern-BioBreaker, a specialized bio-red-teaming model, together with an integrated computational-to-physical framework that couples model-level stress testing with wet-lab validation. Within this framework, Intern-BioBreaker generates targeted jailbreak prompts to test whether aligned models can be induced to provide operational guidance for safety-sensitive biological tasks or produce sequence-level outputs with potentially harmful properties. Selected sequence outputs are then carried forward for DNA synthesis, host expression, and orthogonal protein verification to assess whether model-generated designs can yield the intended biological products. Our evaluation reveals a concerning gap between text-level safeguards and the risks posed by capable scientific models: (i) Intern-BioBreaker outperforms baseline attack models and reveals widespread bio-risk jailbreak vulnerabilities across both open-weight and proprietary frontier LLMs, with several targets reaching near-saturated or 100% task-level attack success rate (ASR); (ii) in sequence-level case studies, GPT-5.5 can be induced to generate modified viral candidate sequences with pathogenic potential; the corresponding translated proteins may exhibit even stronger receptor-binding affinity and thus enhanced infection potential; and (iii) end-to-end verification shows that selected model-generated biological designs are not merely textual artifacts, but can be physically realized under controlled experimental settings. These findings underscore the need for stronger biological red-teaming, nucleic acid synthesis screening, and safety mechanisms that keep pace with model capabilities.

While large language models (LLMs) can solve advanced reasoning problems in seconds, we show that even frontier models fail to perform a much simpler operation: exactly copying an input string that lies well within their context windows. We attribute this failure to positional encodings in Transformer architectures, whose inductive bias favors copying through a shortcut based on matching local contexts rather than carefully locating the corresponding input positions. To address this issue, we introduce 2D-RoPE, which organizes text into a 2D grid rather than a 1D sequence and assigns each token a row ID and a column ID. Under this view, copying becomes simply retrieving input tokens at a fixed column offset, which makes the task easy to learn. In synthetic copy experiments, shallow Transformers with 2D-RoPE achieve perfect copying at input lengths hundreds of times longer than those seen during training, whereas standard positional encodings fall far behind. We further show that the advantage of 2D-RoPE language models on copy tasks consistently holds in large-scale pretraining on DCLM with model sizes up to 1.4B parameters. Overall, our results suggest that viewing text in 2D can benefit language modeling, and we hope this encourages future work to further explore the potential of 2D positional encodings.

The rapid emergence of LLM-powered autonomous and semi-autonomous agents is reshaping software systems from static, request-response components into goal-directed, adaptive, and tool-using computational actors. As these agents move from isolated cognitive prototypes into complex distributed workflows, they confront challenges that the Service-Oriented Computing community has studied for more than two decades: composition, interoperability, quality of service, lifecycle management, governance, security, and trust. Yet much of today's agentic AI ecosystem is developing these foundations ad hoc, without the engineering rigour required for dependable enterprise and societal deployment. This paper introduces Agentic Service-Oriented Computing (ASOC) as a new research and practice area concerned with engineering agents as services, orchestrating services through autonomous and semi-autonomous agents, and governing ecosystems of agents and services under constraints of trust, cybersecurity, compliance, performance, and accountability. We articulate six foundational principles of ASOC (harness-ability, composability, lifecycle engineering, trustworthiness by design, goal-driven orchestration, and observability/accountability) and organise a five-dimensional research agenda spanning: (i) agentic services foundations and lifecycle engineering; (ii) composition, orchestration, and interoperability; (iii) governance, observability, and accountability; (iv) security, trust, and risk management; and (v) evaluation, certification, and Agentic QoS. We argue that the Services Computing community is especially well positioned to provide the conceptual and engineering spine for this emerging field, transforming agentic AI from fragmented demonstrations into dependable, service-based systems worthy of human and organisational trust.

Molecular property models are commonly evaluated by holding out Bemis--Murcko scaffolds, yet a scaffold identifier is only one notion of chemical unfamiliarity. We introduce a label-free structural-frontier split that reserves the sparsest and most physicochemically remote scaffold groups, and evaluate it on six public experimental or curated ADMET tasks. Against a 70/10/20 scaffold control with identical acyclic grouping, the frontier inflates equally weighted primary error with a taskwise median of 87.0% and a skew-sensitive mean of 130.3% (descriptive task/seed bootstrap interval, 52.1--246.0%). The mean falls to 75.9% once BBB is removed; that endpoint is the one whose score ranking inverts at the frontier. A message-passing graph-network control still shows a large gap (mean 82.8% over four tasks) and does not invert, so a low-capacity head does not explain the effect. We also test Multi-View Frontier Risk Extrapolation (\method), a count-adjusted tail-risk penalty over four molecular views, and treat it as a falsifiable probe. It changes normalized frontier error by only 0.16% relative to empirical risk minimization for the perceptron head (interval, $-0.43$--0.84%) and by $-1.9$% for the graph network; three fixed robust-penalty controls are likewise inconclusive. Against the published Lo-Hi and DataSAIL splitters the frontier inflates error more on average, though no split is uniformly hardest. An audit of 31,561 marine natural products further shows that OOD status and agreement with legacy ADMET predictions depend on the molecular view, endpoint and teacher coverage. Split construction and label provenance are important evaluation constraints in their own right, and the tested training penalties do not resolve the frontier failures we observe.

Archive-based exploration methods such as Go-Explore select which visited state to return to using visitation rarity, and frontier methods return to the boundary of the unknown; neither asks whether the unexplored region behind a boundary is enterable at all. Exploration is not just about finding reward - it is about collecting a structurally complete experience for downstream learning and planning. We introduce TopoExplore, which augments Go-Explore cell selection with a periodic topological pass: enclosed unexplored regions (voids) of the visited-set occupancy grid are detected by flood fill (the H1 classes of its cubical complex), and a decaying selection bonus is placed only on their strict entrances (gap or door cells), so sealed regions are never targeted and entered regions retire. On a controlled 18-environment MiniGrid suite (15 seeds, frozen hyperparameters) TopoExplore attains a 1.52x geometric-mean speedup in median steps-to-first-entry over its exact Go-Explore ablation, versus 1.37x for a frontier baseline; frontier exploration degrades when sealed decoy structure appears (0.83-1.48x on decoy environments vs. 1.65-2.11x for TopoExplore), while TopoExplore holds its largest win on hard multi-interaction doors (10.9x). We report an honest negative on Montezuma's Revenge - without wall knowledge, unreachable occupancy artifacts capture the bonus and performance degrades as it grows, isolating the wall-aware entrance test as the load-bearing component - and a preliminary positive on HM3D scanned buildings, where the speedup over Go-Explore tracks scene difficulty (r=0.69) even as frontier selection dominates blanket coverage. The evidence supports a deliberately scoped claim: topology-aware selection pays off where enclosed structure must be discriminated, and remains competitive at open coverage, where frontier methods are strongest, despite not being tuned for that regime.

Masculinity in nineteenth-century fiction is not a single ideal but a field of competing scripts. Drawing on 150 British and American canonical novels from the txtLAB Novel450 corpus, published between 1771 and 1930, this paper examines the changing relative prominence of competing models of masculine authority. To focus the analysis on masculine characterisation, the study extracts male-character-centred text windows by using coreference resolution to group names, nominal mentions, and pronouns into character-specific reference chains. It then fits an unsupervised structural topic model with publication year and author gender as topic-prevalence covariates. The model identifies six distinct masculine formations: aristocratic-chivalric, Christian manhood, gentlemanly respectability, country squire, professional-commercial, and imperial/adventure. Across the corpus, formations tied to inherited rank and sacred authority decline, while those organised around paid work and adventure rise. The largest increase occurs not in professional-commercial breadwinning but in imperial/adventure masculinity, particularly the frontier-wilderness register. The trajectory points to a reallocation from inherited and sacred status towards achieved, commercial, and expansionary forms of masculine authority. Adventurous and commercial formations are also more prevalent in novels by authors recorded as male. Because these formations emerge without a seeded vocabulary yet align with categories established in independent scholarship, the article offers a reproducible method for measuring the reorganisation of gendered authority across the long nineteenth century.

Humans and animals exhibit remarkable robustness in physical manipulation, yet robots remain far behind. Progress toward human-level manipulation robustness is hindered by the absence of a unified and systematic understanding: different subfields frame robustness in distinct ways, often leaving the concept ambiguous and limiting deeper analysis as well as communication across research areas. This paper presents a systematic study of manipulation robustness. We begin with a formal definition, characterizing robustness as the degree to which a manipulation system can achieve its goal in the presence of uncertainty and variation. Building on this definition, we introduce general formulations of manipulation robustness from probabilistic and control-theoretic perspectives. We then synthesize the guiding principles and concrete mechanisms of manipulation robustness across perception, planning, control, policy learning, and hardware, illustrating each mechanism through representative works, including foundational and recent studies. In addition, we revisit existing metrics and evaluation methods for quantifying manipulation robustness. Finally, we distill broader lessons for designing robust manipulation systems and discuss open problems and future directions toward achieving human-level robustness in robotic manipulation.

Cybersecurity is a real-life test-bed for many machine learning problems at once, especially when considering modern strides in using Large Language Models (LLMs) to automate processes as ``agents.'' Cybersecurity workflows require orchestrating hundreds of standard and bespoke tools through various formats. The scale of cybersecurity data is enormous; for example, a single malware sample can be viewed as a sequence of billions of tokens. The cost of labeling any file by experts is enormous and labor-intensive, in part because an adversary (possibly a well-funded nation state actor) is attempting to subvert your detection methods. Even skilled experts may disagree on the correct label, creating ambiguity in what constitutes ground truth. When deployed, models must run quickly on billions of items a day, where low-latency is critical for operational success, in a continuously changing environment. In addition, explainability is not optional: analysts demand clear reasoning for model decisions to cope with the large number of false-positive alerts they face daily, and to quickly develop remediation and understand how something went wrong. In short, the amount of complexity cybersecurity is greater than that of natural language and computer vision, and thus we posit that cybersecurity is the better test-case for general AI progress than other, well-studied fields.

Existing benchmarks typically report accuracy for a single model on a single run. This systematically understates real-world LLM capabilities, particularly under heterogeneous data distributions: (i) different models get different questions correct according to their specializations, and (ii) given a budget, multiple generations can be sampled and selectively retained. To quantify this gap, we introduce the Capability Frontier: a Pareto frontier over a set of models that characterizes the best achievable performance at each cost level under optimal selection across models and generations (i.e., via an oracle). Our construction corrects for two opposing biases: underestimation from single-model evaluation and overestimation from taking maxima over noisy samples. We study 21 LLMs across 16 widely used benchmarks spanning coding, reasoning, medicine, factuality, instruction following, and agentic tasks, comparing Capability Frontier performance at matched cost to each benchmark's top-performing model. Correcting for single-model evaluation yields a 54% error rate reduction; additionally correcting for single runs yields an 82% improvement, with SOTA accuracy matched at 85% cost reduction. Complementing these empirical results, we use controlled probabilistic simulations to show that higher query topic entropy produces a near-monotonic increase in the performance gap between oracle routing and the best single model. Our findings suggest collective LLM capabilities are substantially underestimated, with implications for evaluation and deployment in data-heterogeneous, multi-domain settings.

Image generative models are trained on massive, largely uncurated internet-scale datasets that contain undesirable visual concepts. Efficiently removing such concepts from the model generations without degrading the quality of output images remains challenging. We introduce a novel concept removal method for frontier diffusion and image autoregressive models, such as SD3.5, Flux, and Infinity. Our intervention replaces the internal bottleneck layer present in all these modern models with a transcoder that is trained to replicate the original layer while structuring it into distinct activation features. This in-place substitution creates an integrated filter through which concept-specific signals can be selectively disabled while preserving the rest of the model's behavior. Since the intervention modifies the model backbone rather than attaching an external component, it remains persistent under white-box access. Empirically, the approach achieves state-of-the-art concept removal performance across modern diffusion and autoregressive models, maintains visual generation quality, provides robustness against adversarial prompts, and supports sequential removal of diverse concepts. This positions our method as a practical approach for concept removal in frontier image generative models.

Long-form serialized audio drama, with arcs that run for 200 to 800 episodes, is a major creative medium and a setting where frontier large language models (LLMs) fail. We benchmark 21 models, spanning classical, fine-tuned, open-frontier, closed-frontier, and reasoning tiers, on a uniform set of structural narrative metrics. All closed-frontier systems saturate at a plot-beat F1 in the band [0.78, 0.81] and collapse by about -0.20 F1 at horizon h=200. We introduce NarrativeWorldBench, an open benchmark of nine narrative-structure metrics evaluated across horizons h in {10, 20, 50, 100, 200}, with cross-lingual evaluation across four Indic languages (Hindi, Tamil, Telugu, Marathi). We introduce N-VSSM, a Narrative Variational State-Space Model that maintains a structured 256-dimensional latent world state over more than 200 episodes via a Mamba-2 backbone with an event-conditioned posterior and an 8B decoder. N-VSSM holds plot-beat F1 >= 0.84 across all horizons at 4x lower compute than the closed-frontier band. A learned Cultural Transfer Function lifts cross-language fidelity by +0.20 to +0.23 Likert points. In a within-subjects writer study (n = 12 professional authors, 240 trials), N-VSSM is preferred over Claude Opus 4.5 on long-arc consistency 71% of the time and rated +1.3 Likert points higher on controllability.

Frontier scientific reasoning remains a major challenge for large language models (LLMs), where even the strongest commercial systems fall short of expert-level performance. A closer look at model behavior reveals substantial complementarity that single-model evaluation hides: different frontier models excel on different question types, and no single model captures the full picture. We present SciOrch, a framework that trains a lightweight 8B model to orchestrate frontier LLMs for scientific reasoning. The orchestrator decomposes each question, delegates sub-problems to selected commercial models through API calls, and synthesizes a final answer. Training such an orchestrator is fundamentally harder than conventional agentic RL: each action triggers an API call that is expensive in both dollar cost and latency, making standard online rollouts infeasible. We address this with MCTS-based approach, producing diverse orchestration trajectories, extracting per-node single-turn samples, and optimizing the orchestrator with GRPO-style training. On a 240-question test set spanning SGI-Reasoning and Scientists' First Exam, SciOrch reaches 56.66% average accuracy, outperforming the strongest single commercial model by 3.74% and the strongest multi-agent baseline by 3.33%. It also attains the best accuracy on both SGI and SFE with less than half the API cost of typical multi-agent methods.

This paper provides identification results to characterize a fairness-accuracy (FA) frontier, and statistical inference tools to test hypotheses and build a confidence set for the FA-frontier, when outcomes are observed only for selected individuals. When the selection process is unrestricted but loss is measured in specific ways, we provide a characterization of the sharp identification region of the FA-frontier. Under an assumption of unconfoundedness conditional on observables (and unrestricted loss functions), we obtain point identification and propose a debiased machine learning estimator, derive its asymptotic distribution, and show how this can be used to carry out inference for the FA-frontier. In work in progress, we extend the partial identification results to a broader class of loss functions.

Environmental scientists spend disproportionate effort on data wrangling rather than analysis. New AI agents can be a helpful tool, but no benchmark exists to evaluate AI agents that automate environmental geospatial workflows through structured tool calling against real APIs. We introduce the GeoNatureAgent Benchmark, the first benchmark for environmental analysis agents that operate via structured tool calls to a production-style geospatial API. The benchmark comprises 93 tasks across 18 categories. Tasks are evaluated against an open, self-hostable geospatial API that serves three environmental indicators across Spain and Portugal via sixteen tools. We evaluate nine frontier and open-weight LLMs, reporting capability and per-case cost as orthogonal axes. Results manifest that (1) Claude Sonnet 4 achieves the highest capability at 60.8% +/- 0.8%, followed closely by DeepSeek V3.2 at 56.3% +/- 3.1%, while no other model exceeds 51%; (2) the cost-accuracy Pareto frontier is occupied mostly by open-weight models, with DeepSeek V3.2 offering 93% of Claude's capability at 11.6x lower cost; and (3) structured tool calling against a real API provides a more discriminative measure of real-world agent capability, with mean accuracies 25-35 percentage points below those reported on general-purpose GIS benchmarks.

Vision-Language Models (VLMs) are increasingly deployed as high-level planners for embodied agents, with an emerging strategy of scaling test-time compute to improve capability. However, we observe that doing so increases latency, token usage, and FLOPs while yielding uneven, often diminishing gains in downstream success, limiting where embodied agents can be deployed. We argue that choosing when and where to spend test-time compute is central to bringing frontier performance to the real world. We introduce DIRECT, a routing framework that uses multimodal scene context to allocate compute per prompt, improving the success--cost Pareto frontier over fixed model selection. Across three dominant scaling axes, namely chain-of-thought depth, model size, and memory history, our experiments on VLABench and RoboMME show that test-time compute is not a uniform lever: different axes yield qualitatively distinct capability gains. We validate these insights on a physical Franka arm in a DROID setup spanning zero-shot manipulation and long-horizon chaining, where our router matches or exceeds a stronger model's success rate at up to 65% lower average latency. Ultimately, our results show that naively scaling test-time compute is wasteful, and that DIRECT can provide frontier-level embodied planning in robotic systems at a fraction of the cost. Project page can be found at jadee-dao.github.io/direct/.

The deployment of Large Language Model (LLM) agents for computer automation is accelerating, yet their ability to navigate complex, professional-grade productivity software is largely untested. We argue that Office automation is an ideal environment for benchmarking document-automation capability, as it requires long-horizon planning and reasoning, precise parameter configuration, and multi-application integration. To quantify this capability, we introduce an evaluation based on China's National Computer Rank Examination (NCRE), featuring 200 comprehensive practical-operation tasks across Word, Excel, and PowerPoint. Each task is scored on a 100-point rubric scale using 7,118 machine-gradable criteria, and Score Rate (SR) denotes the mean percentage of rubric points earned across these tasks. We benchmark 7 frontier LLMs and observe stark limitations: single-turn models score a maximum of 36.6%. A stronger agentic system with execution feedback, iterative repair, and broader Office automation access reaches 68.8%, but remains below the 95.5% community-reference score used as a scoring sanity check. Ultimately, our experiments demonstrate that despite recent advancements in code generation, achieving reliable fine-grained Office document automation remains a significant challenge for current code-generating LLM and agent systems.

Heuristics play a central role in the performance of bidirectional search algorithms, which commonly rely on two main classes. Front-to-end (F2E) heuristics estimate the distance from a state s to the target of the search (the goal for forward search or the start for backward search). In contrast, front-to-front (F2F) heuristics estimate the distance from s to the opposite search frontier using a pairwise function h(s, s'), where s' ranges over frontier states. Although F2F heuristics are typically more informative and therefore reduce the number of node expansions, their reliance on extensive pairwise evaluations incurs substantial computational overhead. To address this limitation, we introduce a new heuristic class, front-to-attractors (F2A), that preserves much of the informativeness of F2F while dramatically reducing its computational cost. Rather than evaluating distances to all states on the opposite frontier, F2A estimates the distance from s to a small, dynamically maintained set of attractors in the opposite search direction. These attractors serve as a surrogate for the full frontier, enabling rich heuristic guidance at a fraction of the computational expense while maintaining the optimality guarantees offered by F2F. We evaluate F2A across multiple domains and show that it reduces the number of pairwise evaluations by up to 11.2x compared to F2F, while achieving 4.8x fewer node expansions than F2E on average.

Diffusion Large Language Models (dLLMs) refine tokens iteratively but commit them irreversibly, leading to a "stability lag" where early decisions remain fragile even after being written. We reveal that Post-Training Quantization (PTQ) error easily flips these borderline decisions at the write frontier, which are then permanently locked in and amplified. To address this, we propose Frontier-Aware Instability-Reweighted Calibration (FAIR-Calib), a two-stage PTQ framework for dLLMs. Stage I probes a full-precision teacher to estimate a position prior that combines frontier hits and masked-stage reliability. Stage II performs off-policy, layer-wise calibration by minimizing a reweighted hidden-state MSE, effectively prioritizing the protection of fragile frontier states without requiring expensive end-to-end diffusion rollouts. We further theoretically justify our weighted objective as a surrogate for output KL divergence. Empirically, FAIR-Calib consistently outperforms state-of-the-art baselines on LLaDA and Dream (W4A4), significantly reducing frontier decision flips and suppressing post-commit mismatches across diverse benchmarks.

We introduce PaddleOCR-VL-1.6, an upgraded compact document parsing model built upon PaddleOCR-VL-1.5. Although PaddleOCR-VL-1.5 establishes a strong 0.9B baseline, its remaining errors concentrate in under-optimized regions where model behavior is unstable, data coverage is sparse, or supervision is unreliable. Rather than expanding the training corpus indiscriminately, PaddleOCR-VL-1.6 introduces a region-aware data optimization framework that identifies weak regions from the previous model, applies targeted enhancement to these regions, and improves the reliability of supervision signals. It further adopts a progressive post-training recipe based on curated data selection and reinforcement learning, pushing model performance to a higher level through staged optimization. PaddleOCR-VL-1.6 achieves a new state-of-the-art score of 96.33% on OmniDocBench v1.6, demonstrates strong competitiveness against top-tier VLMs, and provides a practical post-training recipe for the PaddleOCR-VL series.

We describe our submission to the VRR Challenge @ CVPR 2026, built on the \emph{ImplicitQA} / \emph{VRR-QA} benchmark~\cite{implicitqa}: multiple-choice video question answering in which answers are deliberately \emph{not} observable in any single frame and must be inferred from spatial layout, motion, depth, viewpoint, causality, and social context across discontinuous frames of creative video. We conduct a systematic, training-free study spanning open-source Video-LMMs (Qwen2.5-VL~\cite{qwen25vl}, Qwen3-VL~\cite{qwen3vl}, InternVL3, Gemma-3, and the RL-tuned video reasoners Video-R1~\cite{videor1} and VideoChat-R1.5~\cite{videochatr15}) and a battery of inference-time strategies (chain-of-thought, question decomposition, describe-then-reason cascades, audio transcripts, spatial state prompting, self-consistency~\cite{selfconsistency}, multi-model ensembling, and category routing). Our central finding is that this benchmark is \emph{perception-bound rather than reasoning-bound}: reasoning-side augmentations are neutral-to-harmful, whereas base-model perceptual capability and lightweight test-time denoising are the only reliable levers. A per-category error analysis localizes the difficulty to low-level perception -- relative depth, viewpoint, and counting are the hardest categories, while causal and social reasoning are nearly solved -- and a prompt that explicitly injects monocular depth cues to attack the weakest category \emph{lowers} test accuracy by $5.8$ points, confirming that the model needs a better \emph{percept}, not a better \emph{procedure}.

With rapid development of large language models and diffusion-based content generation, world modeling has attracted increasing research attention, benefiting various downstream domains such as game engines, embodied AI, autonomous driving, etc. Through explicitly incorporating user actions into world state transition, recent literature empowers world modeling with interactivity in an action-conditioned video or 3D generation paradigm, further enhancing controllability over world evolutions and facilitating users to freely traverse, manipulate, navigate, and personalize the state evolution. In this paper, we aim to systematically review recent research trends, technical developments, evaluation benchmarks, and also propose future potential directions in interactive world modeling. Specifically, we first summarize recent efforts and trends in terms of application scenarios, world state evolution, and scene modality. Afterwards, we delve into three crucial technical challenges, including action-conditioned controllability, long-horizon interactions and memory, and action-following responsiveness for real-time interactivity. Furthermore, we also thoroughly compare existing benchmarks and metrics in four specific application fields: open-world exploration, game engine, autonomous driving, and robotics. Finally, we discuss several promising future directions in achieving next-generation interactive world modeling. The corresponding repository is publicly available at: https://github.com/liujiuming123/Awesome-Interactive-World-Model.

While continual pretraining~(CPT) is a practical way to extend large language models to new languages, naïve finetuning on targeted data erodes existing capabilities through catastrophic forgetting. Organizing training around language families reduces cross-language interference but cannot alone prevent forgetting of the general knowledge needed for downstream tasks. We link this forgetting to parameter drift in multilingual CPT and present a suite of five layer-aware parameter alignment strategies: hard layer freezing, soft regularization, post-hoc weight reversion, and model merging. We systematically compare our alignment strategies against two unregularized CPT baselines on benchmarks spanning 32 training languages from five language families, plus held-out languages, across four evaluation axes: perplexity, reading comprehension, physical reasoning, and translation. Parameter alignment substantially reduces forgetting at minimal cost to language acquisition: layer freezing and regularization best preserve comprehension, whereas post-hoc reversion yields the strongest translation gains. Together, these results map the acquisition--forgetting frontier for family-expert CPT and offer practical deployment guidelines pairing each strategy to the tasks it best serves.

When a learner faces a new task with few samples, it must leverage any available side information. In practice, this often comes in the form of model evaluations on related tasks in public benchmarks. A key question then is how to model task relatedness such that it is both realistic and the benchmark evaluations lead to provable gains. Empirically, we observe that weak monotonicity is often approximately satisfied: if a model dominates another on many benchmarks, it also tends to outperform on the new task. We explore the statistical complexity of learning under (approximate) weak monotonicity, leveraging it within two learning paradigms: transfer learning and model selection aggregation. We show that not only can we prune the model class based on monotonicity, but we can also further adapt to the geometry of the available trade-offs by hedging on the frontier.

Existing LLM moral benchmarks usually ask which isolated moral act, value, or foundation a model prefers. This is useful but incomplete. Realistic judgments often require a model to combine several moral signals within the same option. We introduce Moral Trolley Arena, a two-stage blind ELO benchmark for measuring how LLMs compose moral evidence. The single-scene arena first calibrates individual moral acts from a 229-scenario corpus across five Moral Foundations Theory foundations; the composite arena then combines calibrated acts into two-act moral items over a controlled intensity grid and measures the resulting composite preferences. Across ten frontier models, composite judgments are largely predicted by component act strength, but the relation is consistently compressed rather than simply additive. Models also show non-additive intensity anchoring, bounded foundation-specific residuals after component control, and highly convergent composite preference surfaces across providers. These results suggest that moral audits should measure composition rules for moral evidence, not only rankings over isolated acts.

We benchmark three supervised fine-tuned models against frontier zero-shot baselines on a 661-row held-out slice of PiSAR (Persona, intent, Screen, Action, Rationale), a 12,929-tuple corpus of screen-anchored behavioural rationales curated from public app-store reviews, Pew American Trends Panel demographics, and the OPeRA shopper traces. Every model, frontier or fine-tuned, is evaluated on the same 661-row slice with the same scoring pipeline. Two findings. First, frontier zero-shot baselines (Claude Opus 4.7 and GPT-5.5) reach sem_sim 0.459 and 0.482 respectively; a fine-tuned Qwen3-VL-8B-Instruct reaches 0.783 and clears sem_sim >= 0.7 on 79% of rows, against 1-2% for either frontier baseline, a gap of 0.30 absolute on the same test set. Second, the same training data and recipe on Gemma-4-26B-A4B-IT scores only 0.441, in the same band as the frontier zero-shot baselines rather than the fine-tuned Qwen. We read this as a recipe-vs-model mismatch: the reasoning-tuned high-parameter model resists displacement and would likely need either more data or a stronger fine-tuning method.

Linking free-text phenotype descriptions to ontology terms, typically referred to as phenotype annotation, is essential for the cross-study integration of comparative morphological data. This labor intensive process has heavily relied on highly trained human experts, which makes it challenging to scale and thus a key bottleneck. Dahdul et al. (2018) established a Gold Standard (GS) of Entity-Quality (EQ) annotations across seven phylogenetic studies and used it to evaluate three human curators and the Semantic CharaParser NLP tool with ontology-based semantic similarity metrics; they reported that machine-human consistency was significantly lower than inter-curator (human-human) consistency. Here we revisit that benchmark with five frontier hosted LLMs from Anthropic and OpenAI, each operating as an "agentic curator" within a self-contained workspace that supplies the source publication PDF, the same annotation guide used by the original human curators, the four project ontologies (UBERON, PATO, BSPO, GO), and a validation script. Evaluated against the same Gold Standard, every agent fell within the range of inter-curator variability of the three trained human biocurators of the original study; the best performing agents approached but did not reach the best performing human curator. Agents substantially outperformed Semantic CharaParser on all four metrics.