Objectives

Goal-conditioned reinforcement learning aims to learn policies that reach specified goals, but remains challenging in offline settings with sparse rewards and long-horizon dependencies. In such settings, goal-completion information can be temporally distant from the early decisions that enable success, while offline value estimation introduces additional error. We study this issue from a reward-propagation perspective and show, in a stylized delayed-goal setting, how goal-directed value separation can become small relative to local estimation error. Motivated by this analysis, we propose Reward Stimulation Implicit Q-Learning (RSIQL), a simple non-hierarchical method that introduces additional reward signals at progress-making intermediate states in offline trajectories. RSIQL uses an auxiliary goal-conditioned value function to identify intermediate states estimated to make progress toward the goal and applies reward stimulation to provide less-delayed training supervision. Unlike hierarchical methods, RSIQL does not learn a separate high-level subgoal policy. Experiments on D4RL goal-reaching benchmarks and OGBench show that RSIQL improves over goal-conditioned IQL on average and achieves performance competitive with hierarchical offline goal-conditioned methods, while retaining a simple flat policy structure.

Two dominant tokenisation algorithms are used by modern language models: byte-pair encoding (BPE) and UnigramLM. These differ along two orthogonal axes: their optimisation objective (compression vs. log-likelihood) and their search procedure (bottom-up merging vs. top-down pruning). Existing comparisons confound these axes, making it unclear whether their observed differences stem from what is being optimised vs. how it is being optimised. We disentangle the two by introducing two new tokenisation algorithms that complete this 2x2 design space: BottomUpLL, a bottom-up likelihood-based tokeniser, and TopDownComp, a top-down compression-based tokeniser. We train language models with tokenisers produced by each algorithm, varying: model size, vocabulary sizes, and domain (English-only vs. multilingual). Evaluating models on bits-per-byte, we find that the search procedure -- not the objective -- is the dominant factor: bottom-up tokenisers consistently achieve lower bits-per-byte in most settings. Evaluating models on the BLiMP task, however, shows no consistent relationship between design choice and performance. Overall, our results disentangle the effect of tokeniser design choices on language modelling performance, offering concrete guidance for their more principled construction.

In this paper, we study the detection of an echo chamber in a social network, i.e., the identification of a set of nodes that agree on a topic, while disagreeing with the rest of nodes. We argue that this problem is different from other social network analysis problems such as community detection, and from other graph problems such as maximum graph cut and maximum clique. To the best of our knowledge, we are the first to formalize the objective function of echo chamber detection, by using the theory of Fourier transforms of set functions (Stobbe and Krause, 2012). We propose scalable semidefinite relaxation, solved via an interior point method and sparse linear algebra. Experimentally, our algorithm recovers the ground truth echo chamber better than competing methods on small synthetic experiments. Our algorithm produces echo chambers with better network properties than competing methods on large real-world datasets. To independently validate our proposed objective function, we show that our algorithm finds echo chambers with more agreements with suspended users than competing methods on a small real-world dataset.

Systematic generalization, the ability to solve novel problems by recombining known atomic elements, is central to human intelligence but difficult to study rigorously under controlled settings. Existing studies therefore rely on simplifications such as approximately linear action composition, productivity-based tests, and action-explicit goals, which make systematic generalization easier to study but omit some essential aspects of this capability. To characterize what these simplifications miss, we adopt a reasoning-centered lens and introduce TranSGrid, a testbed that brings deductive, inductive, and abductive reasoning together within a unified task. Experiments with seven Transformers on 4,800 TranSGrid instances show that all models perform much worse on TranSGrid than on a held-out test set: the largest model solves 79.6% of the test set, but only 55.3% of TranSGrid and 15.8% of the hardest subset. The gap remains within the training length range, showing that productivity alone is not sufficient to evaluate systematic generalization. Additionally, we reintroduce the other two simplifications into TranSGrid: one variant makes actions compose almost linearly (reducing the inductive demand), the other makes goals action-explicit (reducing the abductive one). In both, solve rates return to roughly the test set level, showing that either simplification alone is enough to reduce TranSGrid to an ordinary held-out test set. Together, our results show that existing tasks reduce either or both of the inductive and abductive demands, and that comprehensively measuring systematic generalization requires a task that involves all three forms of reasoning.

Self-supervised learning (SSL) can reduce the need for labelled medical images, but the choice of pretext objective remains unclear for lung ultrasound (LUS). Contrastive learning, masked reconstruction, and joint-embedding predictive architectures (JEPA) differ in the space in which their targets are defined, yet existing ultrasound studies compare them under different corpora, backbones, and evaluation protocols. We compare these three objective families using the same encoder backbone, pretraining corpus, optimisation schedule, and frozen-evaluation protocol. Encoders are pretrained on COVID-BLUeS LUS videos and evaluated with linear, $k$NN, and attentive probes at 5%, 10%, 50%, and 100% label budgets. Evaluation is performed on POCUS using patient-level five-fold cross-validation and on the independently acquired Mendeley-Uganda dataset, which is excluded from both pretraining and probe fitting. At the full label budget under linear probing, VideoMAE and V-JEPA achieve $66.5 \pm 13.1$ and $65.4 \pm 11.7$ balanced accuracy on POCUS, while MoCo achieves $42.1 \pm 1.2$. On Mendeley-Uganda, the ranking reverses: MoCo performs best at $62.7 \pm 1.0$, followed by VideoMAE at $53.8 \pm 2.8$, while V-JEPA falls near chance at $35.1 \pm 4.9$. These results show that POCUS probe accuracy alone does not identify the objective that transfers best across datasets. We also outline planned representation-level analyses to examine this reversal. Code is publicly available at https://github.com/moeinheidari7829/LUSVideoSSL.

Long-horizon goal-conditioned reinforcement learning delegates control to a high-level module that proposes subgoals, but existing subgoals are implicit byproducts of value functions or latent actions, tied to the executor that produced them. We study a different object: a route-conditioned order of unavoidable stages that every successful executor must traverse, recoverable from offline trajectories and belonging to none of them. Its defining properties are topological: an unskippable stage is a separating set that every admissible path must cross, and a loop in free space forces a route choice. We read the two by homology in dimensions 0 and 1 over a transport-weighted carrier built from successful trajectories, yielding an enumerable gate set with shell-level certificates; the certified gates are what we call topological necessities. Certified gates enter the decision loop as a recursive topological gate hierarchy. Under a fixed, isomorphic free space, the object survives executor replacement: gates frozen on PointMaze data transfer without retraining to Ant and Humanoid, attaining the highest Humanoid aggregate under a unified interface (96.1), with +36.0 over a map-privileged reference on the multi-route task (p=1.4e-5); the planner saturates PointMaze (100+/-0) and matches or exceeds the strongest baselines on AntMaze (giant +22.9) and Kitchen (+15.8/+12.6).

We study the time-uniform convergence of the raw iterate of standard stochastic gradient descent (SGD) for unconstrained smooth convex objectives. We prove that, under standard noise assumptions, the time-uniform convergence rate gets arbitrarily close to $\sqrt{\log n / n}$ but never reaches it. More specifically, we prove that for every positive, eventually nondecreasing sequence $h$ satisfying $h(n) = o(\sqrt{n})$, a bound of order $h(n)/\sqrt{n}$, holding simultaneously for all $n$ with probability at least $1-α$ and uniformly over the problem class, is achievable if and only if $$\sum_{j = 1}^{\infty} \frac{1}{h(2^j)^2} < \infty.$$ The constructive sufficiency result follows from a dyadic horizon-free schedule together with an additive conditional-restart inequality. The necessity counterpart applies to every deterministic nonnegative schedule and holds even for a one-dimensional analytic smooth convex objective with Gaussian noise.

Real-world Multi-Objective Reinforcement Learning (MORL) often suffers from sparse rewards, reward conflicts, and late-stage reward tug-of-war, causing traditional linear scalarization to experience severe metric oscillations. To address optimization conflicts among multiple objectives in real-world deployment scenarios, we propose Multi-Marginal Preference Optimization (MMPO), a fine-grained framework that intervenes at the data, gradient, and constraint levels rather than relying on coarse-grained global scalarization. Specifically, MMPO performs exposure debiasing to mitigate sparse and biased rewards, applies priority-aware orthogonal projection to decouple conflicting gradients, and introduces self-prompted gradient constraints to prevent dominant objectives from overwhelming weaker ones. Experiments on real-world e-commerce datasets show that MMPO improves training stability and consistently achieves better performance across conflicting metrics. Moreover, it generalizes robustly to broader tasks such as ToolRL and code generation, demonstrating its effectiveness as a practical paradigm for multi-objective alignment.

We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand how scheming arises from the interaction of key factors, such as instrumental goals, environmental affordances, oversight conditions, and perceived consequences. Prior work examines only a small number of scenarios, limiting the ability to isolate how these conditions shape an agent's propensity or capability to scheme. This limited scale and task diversity also restrict coverage of realistic deployment settings and the range of scheming strategies that can be observed. To this end, we introduce SCHEMEARENA, a 400-scenario benchmark for scalable scheming stress testing, constructed through a factorized scenario synthesis framework spanning diverse safety-relevant tool domains, instrumental goals, oversight conditions, and pressure mechanisms. To enable scalable and reliable monitoring, we further propose SCOUT, a scheming monitor that grounds multi-criteria judgments in evidence drawn from agents' reasoning and actions. Across controlled stress tests on five LLM agents, we find that explicit instrumental goals are the strongest driver of scheming propensity. Strategic hints play a distinct role by helping agents translate scheming reasoning into concrete covert behavior. Oversight has mixed effects: in several closed models, action-only monitoring increases scheming, suggesting that partial oversight can act as an optimization constraint rather than a deterrent. CoT is a useful but incomplete monitoring signal: it can reveal latent scheming before execution, yet action-only scheming shows that covert behavior may occur without explicit reasoning evidence. We release the benchmark, code, and monitor at: https://github.com/launchnlp/SchemeArena.

Offline reinforcement learning (RL) must reconcile two competing requirements: policy updates should stay near dataset-supported actions to keep value estimates reliable, yet meaningful gains often require moving beyond the behavior distribution. We develop a geometric view of offline actor updates by modeling policies as a probability manifold endowed with a chosen metric geometry. Under this lens, a broad class of offline actor objectives can be interpreted as a single proximal policy improvement step (SPI), i.e., an implicit discretization of a manifold gradient flow induced by a critic-defined energy. Building on this insight, we propose multi-step proximal policy improvement (MPI), a plug-in refinement mechanism that composes sequential re-centered proximal steps. MPI enables controlled policy improvement beyond dataset support while retaining proximal control at each refinement. The framework accommodates multiple policy geometries and admits practical instantiations for deterministic and diagonal-Gaussian policies. Experiments on D4RL benchmarks show that small numbers of MPI refinements improve strong offline baselines, including TD3+BC, ReBRAC, and IQL, on many tasks. Focused diagnostics further distinguish re-centered refinement from fixed-objective update scheduling and characterize limitations under critic error.

In an edge--cloud collaborative edge-computing environment, an edge node (EN) must decide whether each user task should be executed locally, forwarded to a remote service (or cloud) node (SN), or rejected. The EN observes its local state directly but receives the SN state only through an intermittently refreshed cache. Status updating and task control therefore form an asynchronous closed loop under partial observability. Freshness-driven schemes, including those based on Age of Information (AoI), do not directly value an update by its effect on subsequent task decisions. We propose CoSMO (Co-design of Semantic-state Management and Offloading), a cooperative event-driven reinforcement learning (RL) framework that coordinates semantic status management and selective offloading through realized task utility. CoSMO learns a compact representation of the heterogeneous SN service state. At the SN, a recurrent semi-Markov double deep Q-network (Double DQN) agent jointly selects send/no-send and the next decision interval. At the EN, a task-terminal off-policy value-learning agent makes hierarchical gate--route decisions from local observations and stale remote semantics. The agents maintain separate observations and value targets but share the same realized task-utility stream, without centralized execution. Across the evaluated workload families, CoSMO's reported relative improvement in on-time completion rate over the best-performing competing method averages 18.6%--21.2%. For capacity-aware decision accuracy across the three strict-overload points, the corresponding reported gains average 17.6%--$17.9%.

Many important forms of human learning begin with a vague goal, such as "become a better physicist" or "improve at research." Learners must interpret the goal, identify capability gaps, decide how to learn, and determine whether they have actually improved. In contrast, existing work on LLM self-evolution typically begins with tasks and evaluation metrics specified by humans, reducing self-evolution to optimizing an explicit objective rather than deciding what and how to learn. We introduce ASPIRE, a benchmark for vague-goal-driven self-evolution. ASPIRE provides only a natural-language capability goal while downstream evaluation tasks remain hidden. The agent must operationalize the goal by choosing data and update methods, constructing training and validation signals, and deciding when to evaluate. ASPIRE supports both model-weight and agent-harness evolution in a unified interactive environment and evaluates the resulting systems on a hidden, expert-authored set of 520 items spanning six goals. Our experiments show that vague goals redirect search effort toward goal interpretation. Current agents routinely complete training and harness-editing loops, but weight-level gains remain sparse and unstable, and the strongest evolved harness remains below the engineered Qwen-Agent reference. Agents often train on mismatched data and trust narrow self-evaluations, so local gains fail to transfer to hidden evaluation and continued search and training can erase earlier improvements.

Psychometric questionnaires contain rich item-level information, yet it remains unclear whether different representation learning objectives recover the same latent organization. We investigated this question using 757 matched teacher-child pairs from the baseline assessment of the Cyprus ProW preschool trial. Behavioral structure was characterized from child SDQ, ASBI, and CBRS item responses using principal component analysis and clustering, yielding four behavioral phenotypes. A contrastive objective substantially improved teacher-child retrieval relative to PCA-based representations, increasing Top-1 accuracy from 0.13% to 7.27% and Top-10 accuracy from 1.98% to 56.14%. However, contrastive representations preserved behavioral phenotype structure less effectively than PCA-based representations. A multi-task objective jointly optimizing alignment and behavioral prediction partially restored behavioral organization but reduced retrieval performance. These findings indicate that teacher-child correspondence and behavioral phenotypes represent distinct forms of latent organization and demonstrate that the latent structure recovered from linked psychometric data depends on the representation learning objective.

Continuous self-improvement requires an ever-expanding pool of self-generated, diverse, adaptive goals. For language agents, existing training environment pools (hand-curated, statically synthesized, or frozen-verifier) keep the goal distribution fixed as the learner scales. We introduce SPADE (Self-Play in Adaptive Synthetic Executable Environments), a self-play RL framework in which a single LLM plays two roles: an Environment Designer that writes complete, long-horizon training environments as executable code with an OpenAI Gym-style reset()/step() interface, and a Reasoning Agent that learns to act in them. Each is a stateful, multi-turn environment (state transitions, reward functions, and verification code), so one interface spans reasoning problems and multi-step agentic tool use. The Reasoning Agent's regret is estimated using the gap between its reward with and without privileged hints; in optimizing this regret signal the Environment Designer learns to target environments at the edge of the agent's capabilities while keeping them feasible. Through extensive experimentation, we find several components critical to success: grounding the Environment Designer on documents sampled from a large pretraining corpus, and giving it an accumulated environment memory. Scaling to 30B-parameter models, SPADE improves over the strongest fixed-environment baseline by +5.3 on average across eight held-out math, science, code, and reasoning benchmarks, and lifts the tool-use setting by +5.7 on BFCL-v4 multi-turn and +13.9 on ACEBench-Agent; on the games setting, the margin over the strongest baseline grows with model scale. By making environment design itself a learnable component, SPADE takes a concrete step toward open-ended self-improvement.

Crowd counting must recover reliable local density under severe variations in perspective, head scale, occlusion, and background clutter. Although modern counting objectives provide strong spatial supervision, many multi-level decoders still use spatially invariant feature fusion and apply one receptive-field pattern to every location. We propose DCA-MoE, a framework that makes both decisions content dependent while retaining a frozen DINOv3 encoder. Spatially Adaptive Layer Fusion (SALF) predicts position-wise weights over four aligned backbone features, and Density-Routed Multi-Receptive-Field Experts (DR-MoE) assigns each location a soft mixture of local, mid-range, and large-context residual experts. An EBC-style head reconstructs block density, while DMCount supervision and an auxiliary routing-balance term train the decoder without updating the backbone. On the NWPU-Crowd validation split, the strongest paired configuration, based on DINOv3 ViT-L/16, obtains 31.7 MAE and 72.2 RMSE; the matched ViT-B/16 full model obtains a paired 32.2/75.9. Cross-dataset results remain mixed, and several component baselines currently report independently selected minima from a single seed. The evidence therefore supports the feasibility of spatially adaptive fusion and routing, while broader paired and multi-seed evaluation remains necessary for causal attribution.

Latent world models are judged by how well they predict, so when planning fails at long horizons the natural reading is that the predictor degrades. On a reproduction of LeWorldModel on TwoRoom we show the binding constraint is the planner's objective instead. The predictor is not the limit: its imagined state seventy-five environment steps ahead is still only 0.189 as wrong as assuming the world froze, while the planner never imagines beyond twenty-five. The objective is. Cross-entropy-method planning minimises squared latent distance, which tracks true distance at r = 0.426, saturates by about eighty arena units and decreases beyond a hundred and twenty, so moving away from the goal can lower the cost. The information is present throughout: a ridge probe recovers position from the frozen embedding at R^2 0.9922. The pathology is the method's, not one reimplementation's. It is present in the authors' released weights, and across four checkpoints long-horizon success rank-orders exactly with metric quality and inversely with prediction accuracy. Replacing only the objective, with nothing retrained and no GPU, lifts goals reached at offset 100 from 26.0% to 98.0%, equals the 98.0% at offset 25, and reaches 92.0% under a third of the budget: planning stops depending on the horizon. The best cost is not the most accurate. A head learned from frame separation alone predicts spatial distance worse than a position probe (r = 0.819 against 0.9897) yet plans better, charging 24% more to cross the environment's dividing wall where squared latent distance charges 4% less. It has learned reachability, not proximity.

We present a formal process to enable non-experts to instantiate and iterate on human-aligned reward functions, i.e. reward functions that adhere to a given preference ordering over trajectories. Given a task described in natural language, our process produces a linear reward function in three steps: distill the task's objectives into a set of fundamental objectives and derive measurable outcome variables that capture those fundamental objectives, select a causally representative subset of outcome variables as the reward terms, and fit weights to those reward terms via preference elicitation. Our contributions describe the first step and formalize the latter two steps. The first is a guided workflow for deriving outcome variables. The second is a reduction of reward term selection to minimum-cost partial cover on a causal DAG, solved in polynomial time via max-flow. The third is a geometric framing of weight fitting as a convex feasibility problem iteratively narrowed by preference queries, solved by existing separation oracle methods. To the best of our knowledge, this is the first reward-design method that maintains a deterministically conflict-free feasible weight region, narrowed to a desired tolerance via a separation oracle with O(n log κ) preference queries.

Is it possible to find an objective classification criterion that organizes the complexity of gust-induced loads across many flight conditions? And one that remains as interpretable as a labelling based on coarse parameters, such as the flight attitude? Our approach encodes a large number of experimental observations through a machine-learned representation and applies a summarization procedure to select a minimal subset of highly significant exemplars. The exemplars provide a similarity-based objective classification criterion of all the observations, they can be more conveniently inspected by experts and can become subject of more refined experiments. We demonstrate the approach on a database of 3480 pressure-load measurements induced by random gusts on a flying-wing model across six flight attitudes. We find nine fundamental response types that recur across multiple attitudes; analysis of a type's transient response enables physical intuition into the underlying fluid mechanics.

Aligned large language models (LLMs) are expected to exhibit safety behavior based on the content of the user request: they should refuse unsafe requests and comply with safe ones. However, we show that the same request can elicit substantially different safety decisions under different traits assigned in the system prompt, a failure mode we call trait-induced safety variation. To measure this failure, we introduce refusal-based metrics: Trait-Induced Deviation measures dataset-level deviation from the no-trait baseline, while Trait-Induced Flip Rate measures whether the same request receives different safety decisions across traits. We then provide a representation-level analysis of the mechanism behind trait-induced safety shifts and find that traits perturb the model's safety representations within a low-dimensional subspace. To achieve trait-invariant safety, where safety behavior remains stable across traits, we introduce Trait-Invariant Safety Tuning (TIST), a simple yet effective self-distillation framework that aligns an LLM's trait-conditioned behavior with its no-trait behavior. Guided by our analysis, we further propose Trait-Subspace Neutralization (TraSN), an instantiation of TIST, which enforces invariance only within the identified trait subspace. Experiments show that TraSN improves trait-invariant safety and strengthens harmful-request safety while preserving general capability. Our results highlight traits as an important factor in LLM safety and robust model behavior.

At corpus sizes typical of low-resource dialects, single-run comparisons can yield gains that do not replicate. We show this for Garhwali, an under-resourced Indo-Aryan language of the central Himalaya, building the first reproducible multi-seed ASR benchmark on the official VAANI splits, with per-seed outputs and significance testing. Re-examining plausible gains, we find them fragile: neither Focal CTC nor a matra-weighted objective beats standard CTC under seed-level testing, the matra objective fails to cut even its targeted errors, and Hindi-to-Garhwali transfer gives no gain over direct fine-tuning. What holds up is mundane: w2v-BERT 2.0 with standard CTC reaches 47.0% WER over five seeds, beating the larger MMS-1B and comparable models; pretraining design, not parameter count, drives performance, and speed augmentation gives a small, largely consistent gain. Multi-seed evaluation on official splits separates real gains from seed noise.

Invariant learning seeks representations that remain predictive across environments, yet the behavior of its objectives along the regularization path is often opaque. We address this objective-behavior gap by viewing representation learning as multimode magnetization and deriving, from concrete invariant-learning objectives, a Landau-type effective free energy whose low-order coefficients form objective signatures and induce distinct regularization phenotypes. Effective quadratic corrections move the phase boundary and enable finite-strength mode elimination; quartic corrections regulate post-onset amplitude and typically leave residual loading at finite strength; higher-order structure governs non-monotone tails, instability, and collapse at large regularization. In a canonical bilinear model, the theory yields closed-form phase boundaries and steady-state loadings, as well as distinct critical strengths for shortcut and stable modes that define a selective-retention window. Controlled experiments confirm the predicted phase boundaries, loadings, and regularization phenotypes. In one- and two-hidden-layer ReLU networks, the same signatures remain predictive of qualitative regularization-path behavior despite depth-dependent shifts in scale. A matrix extension generalizes the framework to coupled collective modes and yields a spectral phase-boundary criterion. Together, the framework turns low-order objective signatures into predictions of regularization phenotypes and, ultimately, of what models learn as regularization varies.

Large-scale learning systems often face the challenge of balancing multiple, potentially competing objectives, such as fairness, accuracy, and latency. While recent work has formalized this as an optimization problem over binary states, many real-world control parameters, such as fairness thresholds, diversity mixing rates, or resource budgets, are continuous. In this work, we extend the framework to \emph{continuous state spaces}. We model the problem as minimizing a sum of linear objectives subject to \emph{movement costs} that penalize system instability. We capture the local structure of the objectives using a \emph{dependency graph} (or factor graph), where each objective is determined by a subset of the state attributes. To address the tension between exploration and stability, we propose \emph{Lazy Graph-LinUCB}, an algorithm that performs lazy updates to minimize switching costs while maintaining near-optimal regret. Beyond stability, we introduce three advanced mechanisms to exploit the underlying graph structure: (1) an \emph{asynchronous} update schedule that eliminates synchronization overhead in sparse graphs; (2) an \emph{adaptive} algorithm that learns the graph structure from data; and (3) a \emph{joint estimator} that leverages data sharing among correlated objectives to significantly tighten regret bounds. Empirically, we demonstrate that these structural exploitations reduce movement costs by more than a factor of three in heterogeneous systems while maintaining similar cumulative losses.

Allocating limited computation among concurrent learning tasks is difficult when each task must reach a target loss before a deadline but its required training effort is unknown. Existing approaches combine online loss prediction with adaptive resource allocation, yet commonly treat computation as continuously divisible throughput. We instead study a practical setting in which tasks arrive over time and computation is provided by discrete nodes. This setting introduces both uncertain demand and constrained sequential decisions. We propose MARA, which predicts future loss trajectories with conditional flow matching and coordinates compute nodes through a cooperative multi-agent autoregressive policy. A potential-based progress reward supplies intermediate training feedback while preserving the undiscounted task-completion objective. Across in-distribution, reinforcement-learning, and vision workloads, flow matching reduces remaining-resource prediction error relative to weighted least squares. At the scheduler's training load, MARA completes 63.46% of tasks on average, 8.54 percentage points above strong baseline Learning with Adaptive Resource Allocation (LARA), and remains ahead under unseen heavier workloads.

Production planning in the manufacturing industry often relies on the use of optimization models, but defining an appropriate objective function can be a challenge. In practice, planners must balance competing goals, manage uncertainty, and account for qualitative business preferences that are difficult to quantify. As a result, many optimization models fail to match expert behavior, limiting trust and adoption. In this work, we propose a data-driven inverse optimization framework to infer the objective function implicitly captured in expert planners' decisions. We formulate the production planning problem as a mixed-integer linear program, where the unknown objective function is represented as a weighted sum of hypothesized cost terms. A suboptimality-loss-based inverse optimization method is then applied to learn the objective weights from historical production plans. The proposed approach is applied to a real industrial case provided by Dow, where the inferred weights reveal that avoiding inventory shortages and maintaining consistent cycle lengths dominate the planners' decision-making. Time- and product-dependent extensions further improve predictive accuracy and uncover evolving priorities. Expert interviews confirm the practical validity of these insights. Overall, this study shows that inverse optimization can transform tacit human expertise into interpretable models, enabling more accurate and trusted decision-support tools for complex industrial systems.

When an AI system is deployed, the individuals who use and or are evaluated by it form beliefs about how the system operates and use those beliefs to strategically present their preferences, behaviors, or attributes. The system then responds with feedback or a decision outcome, thereby creating a human-AI interaction loop. This thesis studies how to design and shape such interactions to achieve three goals: (1) help individuals develop accurate beliefs about the AI systems so they can improve and or secure favorable outcomes at minimal cost, (2) encourage improvement and or discourage gaming behaviors, and (3) ensure that the AI system continues to achieve its intended objectives, such as maximizing accuracy. To address these goals, the thesis is organized into three complementary parts that examine and study human-AI interactions from the perspectives of both evaluated individuals and AI systems. Together, the work presented in this thesis advances human-centered machine learning by providing principles and methods for designing AI systems that align with human needs, values, and capabilities. Methodologically, this thesis integrates theoretical analysis, data-driven modeling, human-subject experiments, and empirical evaluations on real-world and semi-synthetic datasets.

Supervised training of finite-element (FE) surrogate models requires reference solutions, and each reference solution is obtained by solving the system that the surrogate is intended to replace. The assembled discrete potential energy provides a training signal that requires no reference solution. This note records, with proofs, the identities that make this signal exact for linear elastostatics: the difference between the energy of a prediction and the energy of the reference solution equals one half of the squared stiffness-norm error, and the gradient of the energy equals the stiffness-weighted error. Label-free discrete-energy minimisation and supervised regression in the stiffness norm therefore have the same unique minimiser and identical gradients at every point. Around this central result, the note states a conditioning lemma that bounds the displacement error by the energy gap, a modewise contraction identity that explains why the Euclidean displacement error is an unsuitable primary metric, the Chebyshev bound that governs conjugate-gradient post-processing of surrogate predictions, and a conditional latent-separation proposition for joint-embedding predictive architecture (JEPA) pretraining on a shared stiffness operator, with an explicit numerical counterexample that delimits its scope. Every claim with numeric content is implemented as an executable falsification check; the checks were executed twice, on synthetic test problems and on a probe set of 16 instances from the validation split of a pre-registered experimental run, and every inequality holds, with the measured tightness reported. A closing section explains why the construction does not extend to elastodynamics through direct minimisation of the action functional, and which time-discrete formulation restores exactness.

Service robots operating for months in the same homes, offices, and facilities should become more reliable with experience instead of searching familiar space from scratch for every request. Yet ObjectNav is predominantly formulated as one-shot exploration, leaving a central deployment challenge unresolved: recognizing an object does not identify a reachable place to stop, and one confident map error can terminate the task. We introduce SSTG-Nav, a reusable metric-semantic memory that turns a one-time survey into actionable object goals, consolidates evidence across viewpoints, and retains spatially distinct recovery standoffs. On 1,000 HM3D-v2 episodes across 36 scenes, our goal-independent topology achieves a 99.4% geometric success ceiling. Holding semantic responses fixed, metric grounding raises SR/SPL from 0.835/0.560 to 0.920/0.603, and source-aware fusion reaches 0.926/0.586. Fusion-aware Top-3 recovery raises Success@1/2/3 to 0.928/0.965/0.975 and reaches 0.601 SPL@3. Model, field-of-view, density, and corruption controls identify where these gains originate, and a ROS2/Nav2 realization demonstrates the complete reusable query-to-execution pipeline. Together, the results establish pre-exploration as a powerful practical regime for dependable, repeated semantic navigation.

Submodular Information Measures (SIMs) have recently emerged as a powerful framework for representation learning and multimodal learning. In particular, the SCORE framework~\cite{majee2024score} demonstrated that SIMs can serve as effective objectives for supervised contrastive learning. Despite their empirical success, however, the geometric and statistical properties induced by different submodular information measures remain poorly understood. In this work, we develop a unified theoretical framework connecting SIMs to classical concepts in representation learning and statistical pattern recognition. We show that Total Information (TI) objectives characterize intra-class structure: Graph Cut TI recovers within-class variance, LogDet TI recovers generalized variance and covariance volume, and Facility Location TI induces imbalance-aware separation that emphasizes rare and confusable classes. We further show that Mutual Information (MI) objectives capture complementary notions of inter-class structure: Graph Cut MI is closely related to centroid separation and Fisher-style discrimination, LogDet MI captures covariance-aware separation through Mahalanobis distance, and Facility Location MI measures nearest-mode representational overlap. We validate these theoretical characterizations using controlled synthetic experiments that independently vary variance, covariance, class imbalance, class separation, and multimodal overlap. Across all settings, the empirical behavior closely matches the proposed theory. Our results provide the first unified geometric and statistical understanding of submodular information measures and offer principled guidance for selecting and designing SIM-based objectives for representation learning.

Black-box combinatorial optimization requires systematically identifying high-quality solutions under a limited evaluation budget, yet the unknown objective function provides little guidance for deciding where the search should explore next. We introduce SCOPE, a general framework for Synthetic Conditional Objectives for Policy Evolution in Black-Box Combinatorial Optimization. Rather than directly optimizing the inaccessible objective, SCOPE learns a set of synthetic objectives conditioned on the accumulated search history, where each objective is designed to expose a distinct and potentially useful preference over candidate solutions. These objectives are then used to evolve search policies that generate diverse candidates, whose true quality is subsequently assessed through black-box evaluations. The outer loop adaptively updates and selects synthetic objectives according to how effectively their induced policies discover promising regions. In contrast, the inner loop returns a portfolio of top-performing policies to reduce the risk of relying on a single surrogate preference. This formulation reframes objective design as a mechanism for guiding policy exploration, enabling the search process to exploit observed evidence while maintaining structured diversity across discrete solution spaces. Extensive experiments across multiple benchmark problems demonstrate that SCOPE consistently improves black-box search performance under limited evaluation budgets and generalizes well across diverse combinatorial structures.

New venture start-ups need to ``survive'' through multiple stages of reaching milestone targets. We investigate the strategies for start-ups in a milestone-oriented setting. We examine a model of an entrepreneurial start-up firm, where its state is captured by a diffusion process. The entrepreneur can choose between multiple activities (or controls), which incur different cost and determine the drift and the variance of the process. Depending on whether the process reaches a fixed upper boundary or a lower one, the start-up firm succeeds or fails. Continuous-time stochastic models with multiple ($\ge 3$) controls are typically very challenging to deal with. In this work, we are able to completely solve for the optimal policy and provide an explicit characterization of its structure. In particular, the optimal policy only uses controls from a set characterized by a so-called efficient frontier curve that orders the controls by two intuitive measures: riskiness (drift-to-volatility ratio) and cost-effectiveness (drift-to-cost ratio). A unique feature of our model is that depending on the model parameters, the efficient frontier curves can be of different types, resulting in qualitatively different structures of the optimal policy. As far as we know, this is the first study that analyzes a stochastic control model which admits efficient frontier curves of different types. Our work provides start-up firms with intuitive measures to evaluate their activities and offers valuable insights on how the optimal strategies in a milestone-oriented setting change qualitatively contingent upon the specific scenario. We believe the results provide a foundational block in the study of entrepreneurial decision-making.

Good action rankings do not make a contrastive critic safe to maximize. These critics increasingly act as value-like objectives for best-of-$K$ selection, planning, and critic-guided generation. Unbounded bilinear scores can let large embedding norms inflate off-support values, but cosine bounding does not remove the failure. A controlled support decomposition attributes most raw bilinear regret to norm drift. Cosine and hybrid critics nevertheless select off-support actions from most pools and incur comparable regret. Contrastive scores are weakly calibrated or inverted in the top score decile across four OGBench navigation tasks, and they fail to order fixed-query actions by value. Bellman-trained TD-Q succeeds, including in a parameter-matched function-class control. Realized costs depend on the task: simulator rollouts reveal single-step selection costs on PointMaze and the exact-$Q^*$ toy but well-powered nulls on AntMaze and HumanoidMaze, where the controller can self-correct. A training/readout decomposition traces the lost ordering to the cosine training objective; raw-trained embeddings retain weak ordering after inference-time normalization. Candidate maximization can therefore exploit false positives caused by norm drift, score saturation, or in-support misranking. Contrastive critics remain useful compatibility rankers on navigation and manipulation tasks, but action selection requires a value-calibrated scalar.

Large Language Models (LLMs)-powered multi-agent systems are increasingly deployed in mixed-motive environments, where agents operate under asymmetric information and strategic deception due to conflicting or hidden objectives. In these settings, misalignment with collective goals becomes a central concern. We propose a novel framework for evaluating objective misalignment using the social deduction game Werewolf, modifying the objective of a single agent while preserving its assigned role. Across LLMs from four different model families and sizes, four player roles, and three objective formulations, we introduce a dual analysis of the agents' internal reasoning and their public cheap-talk behavior (i.e costless, non-binding communication that does not directly affect the agents' utilities), complemented by an analysis of game outcomes. Our results show that objective misalignment undermines outcomes in inherently adversarial environments, an effect exacerbated by asymmetric information and specialized roles. While compromised agents consistently develop distinct objective-dependent reasoning strategies, these adaptations remain largely invisible in their public behavior. More broadly, our findings suggest that even subtle objective misalignment can profoundly affect collective decision-making, highlighting the need for effective mitigation strategies for LLM-based multi-agent systems.

In some real applications a plan may later become unfeasible due to newly imposed budget constraints, yet, at the same time, using only the original actions of the plan and their order is mandatory. In this paper, we study the problem of extracting, from a precomputed plan, a valid subplan that maximizes utility while respecting a cost bound. Each goal is given a utility value and the plan is reduced by removing actions that support low-utility goals, while preserving both executability and the original action order. We show the decision variant is NP-complete and propose two exact methods to solve it: one via oversubscription planning (OSP) and another via Integer Linear Programming (ILP). This paper extends our previous work published at ICAPS 2026 (Del Toro, Fuentetaja, and García-Olaya 2026b). While the core framework remains as introduced there, we further introduce a refined ILP formulation that significantly decreases the model size and improves computational efficiency.

Even a current high-capability LLM can appear safer when shown a dangerous objective directly than when other agents transform and relay its direction. Using OpenAI's gpt-5.6-sol model alias, we test 25 pre-specified mirrored trade-off profiles. Direct exposure to an objective authorizing concealment, fabrication, and pressure produced advice net opposed to its target. After an Id and Censor transformed the same objective into affect and a constraint-rewritten, target-bearing intention, the user-facing Superego---which saw the preferred direction but not the raw objective, its manipulative clauses, or its source---produced advice net aligned with the target. This behavioral reverse shift is consistent with the model recognizing or distrusting the manipulative motive, although we do not identify its internal mechanism. The second result exposes a compositional safety gap: a current high-capability model can be used as the user-facing component of an automated, multi-stage workflow serving an explicitly manipulative objective. The workflow can keep the raw instruction, its manipulation-authorizing clauses, and its provenance outside the downstream model's context while preserving the objective's target direction. A user with endpoint-only access likewise cannot directly inspect those upstream messages including the objective.

Activation steering provides a lightweight mechanism for controlling large language models at inference time, but steering vectors can have unintended externalities: utility vectors may weaken safety behavior, while refusal vectors may induce over-refusal on benign prompts. We introduce OPIUM (Optimizing Protected Injections via Utility Manifolds), a training-free method for sanitizing steering vectors through representation matching. Given reference behaviors on two prompt sets, OPIUM optimizes a new steering vector that preserves the downstream representations induced by the desired intervention while matching a safer reference behavior on prompts where the original vector fails. Across steering-externality and over-refusal settings, OPIUM improves the safety--utility tradeoff relative to vanilla steering and directional ablation, suggesting that harmful side effects of activation steering can often be mitigated directly in activation space.

We study the coupled objective J_K^WOR = E_{S ~ PL-WOR_K}[max_{i in S} R_i]: the expected maximum reward of a size-K Plackett-Luce draw without replacement, the law of Gumbel-Top-K / Stochastic Beam Search decoding. This estimand differs from the conventional i.i.d. objective J_K^iid = E[max_{i<=K} R_i] targeted by existing sample-reuse Max@K estimators, and reusing their i.i.d. weights under the coupled sampler is provably biased (a closed-form three-item instance gives E[g_iid] = (4/5) grad J_K^WOR exactly; pass@K under the coupled sampler is the binary-reward special case). Generic joint-score REINFORCE is already unbiased for J_K^WOR; what it lacks is sample reuse. Our contribution is to instantiate standard rank-conditioned Horvitz-Thompson estimation for the J_K^WOR subset total: from one Gumbel-Top-n pool (n>K) and its observed priority threshold we build an estimator that reuses all C(n,K) embedded K-subsets, unbiased with an unbiased exact score-function surrogate gradient, plus a reward-sorted Max-specific dynamic program that collapses the C(n,K)-term subset sum (with K!-cost set probabilities) exactly to a one-dimensional integral. A fixed-Q quadrature evaluation costs O(n log n + nKQ) arithmetic and is numerically, not algebraically, exact; no epsilon-approximation rate is certified. Each nonzero degree-K Horvitz-Thompson term has finite second moment exactly when n >= 2K; under the same assumptions the full surrogate gradient has finite second moment whenever n >= 2K (sharpness there is open). At K=1 the construction recovers classical priority sampling. All quantities require only the values and differentiable computation graphs of the n+1 drawn items' probabilities, so finite structured sequence policies sampled by exact SBS are covered. A certified finite-Q quadrature bound and countably infinite support remain open. Validation code is included as ancillary files.

We present a manipulation planning system based on affordance recognition and action effect prediction. The system reasons through possible futures in visual form, and evaluates candidate plans by agreement of predicted outcomes with text-based goals set at run-time, using a multi-modal goal-matching module. Positions of objects named in the goal text are tracked through predictions even when occluded, making it possible to generate action plans even when objects become occluded, or when their initial descriptors cease to identify them in future states. We further expand the system with an image conversion module for translating real-world state images with objects of varied shapes and visual appearances into a consistent visual appearance, to facilitate manipulation planning in a physical robot setup. We evaluate performance of the system's modules in isolation and demonstrate the integrated system's manipulation planning capabilities on a set of challenging tasks in both simulation and on hardware.

Dysgraphia is a specific learning disability that is prevalent among school-age children. It affects handwriting coherence, quality, fluency, and legibility, often hindering academic achievement and early learning development. This motor coordination disorder is typically diagnosed through subjective assessments based on clinician observation, which can be timeconsuming and prone to variability. In this paper, we introduce a deep learning-based framework for objective dysgraphia detection using online handwriting data captured via digitizing tablets. The proposed framework relies on two complementary branches: the first pipeline extracts both handcrafted and embedding-based kinematic features directly from raw temporal signals, while the second leverages image-based representations of the temporal signals generated using continuous wavelet transforms (CWT) and Gramian Angular Fields (GAF). The resulting features are then fused to leverage the complementary strengths of both representations. The four representations were evaluated separately and jointly using the publicly available DiaGraMo dataset, showing that the fusion of GAF, MOMENT, and hand-crafted kinematic features outperforms each individual representation, as well as other fusion schemes. These findings highlight the potential of the complementarity of image and signal based representations for more objective dysgraphia detection.

The article investigates the fundamental problem of ensuring the stability of operator control and preserving goal-targeting in hybrid human-machine decision support systems (DSS) of a new generation. Based on a two-month continuous longitudinal experiment on the joint design of a monograph-format textual array, the latent phenomenon of semantic context drift in large language models of deep logical reasoning (Reasoning LLMs) is verified and described. A mathematical model of interaction in the human-machine interface is proposed, and an original metric is introduced - the operator control stability coefficient, which takes into account the non-linear contextual pressure of hidden reasoning chains. Within the paradigm of the cognitome theory, a critical point of control functions inversion is captured. Engineering recommendations are formulated for implementing dynamic relational arbitration loops based on a modified hierarchical similarity model.

Latent world models enable planning from high-dimensional observations by predicting future states in a compact latent space. However, these models are typically kept frozen at test time: when their predictions become inaccurate, planning can fail, especially under test-time distribution shift. To address this, we propose AdaJEPA, an adaptive latent world model that performs test-time adaptation within the closed loop of model predictive control (MPC). After training, AdaJEPA plans and executes the first action chunk, uses the observed next-state transition as a self-supervised adaptation signal, and replans with the updated model. This closed-loop update continuously recalibrates the world model without additional expert demonstrations. Across a range of goal-reaching tasks, AdaJEPA substantially improves planning success with as few as one gradient step per MPC replanning step.

Two-sided marketplaces connect distinct user groups whose interests often conflict -- improving outcomes on one side could degrade the other side's experience. To address this challenge, we deploy an integrated framework for personalizing free-value thresholds -- a policy governing the scope of complimentary services for job listings -- across a two-sided job marketplace connecting millions of employers and job seekers. Our personalized policy delivers statistically significant and economically sizable lift in the target metric while respecting engagement guardrail constraints. Direct application of standard uplift methods proves insufficient here for two reasons. First, cross-side externalities demand multi-objective optimization: maximizing employer-side metrics risks harming job seeker engagement, with effects varying substantially across job segments. Second, marketplace interference necessitates cluster-level randomization, limiting us to few discrete treatment levels -- effectively a form of positivity violation that rules out methods designed for continuous treatments. We contribute an integrated framework with three components. Our ensemble-based hybrid ranking models target and guardrail metrics separately, cutting guardrail risk by over 10% for equivalent target gains compared to single-objective approaches. A treatment effect extrapolation method extends our estimates from limited experimental variation to untested policy levels, relying on monotonicity assumptions that we validate empirically. Finally, we present production deployment, where post-launch data confirms both extrapolation accuracy and guardrail compliance. Our deployed system demonstrates that principled methodology can enable meaningful personalization even when experiments are severely constrained and different objectives compete -- common conditions that characterize many real-world marketplaces.

Joint-embedding predictive (JEPA-style) objectives learn representations by predicting future latents. In doing so they can discard features that are exogenous (uncontrollable by the agent) yet control-relevant, even when those features are trivially encodable. This occurs because the objective optimizes temporal predictability rather than control-relevance. We isolate this failure mode in a controlled 2x2 experimental design that varies feature controllability and relevance independently, using a predictability knob that decouples a feature's temporal predictability from its control-relevance. Comparing six objectives: reconstruction, JEPA, action-conditioned JEPA, controllability-based JEPA, inverse dynamics under a random policy, and reward-grounded JEPA, we observe that all evaluated reward-free predictive objectives leave the exogenous control-relevant feature near chance accuracy, while a reward-grounded variant retains it selectively. The remedy is label-efficient and robust: as little as 2% of reward-labeled transitions recovers the feature, the effect holds across two environments with different surface forms, and it persists across latent dimensions from 16 to 1024. Comparing the learned latent geometry against bisimulation theory's prediction, the JEPA latent realizes only a small fraction of the class separation a supervised reference attains.

Deep learning problems rarely involve objectives that are equal in importance. A primary objective defines the goal, whilst secondary objectives, such as sparsity, compression, or robustness constrain the solution. While existing multi-objective methods have proven effective in practice, they have a clear symmetry problem and neglect the inherent objective hierarchy built into these objective spaces. We introduce Priority-Constrained Descent (PCD), a gradient-based optimization framework designed to explicitly exploit hierarchical objective structures. PCD preserves the direction of primary descent whilst allowing for the minimal distortion necessary to guarantee progress on secondary objectives, controlled by a single $τ\in [0, 1]$ that dictates the strength of the distortion. The resulting formulation is invariant to objective scaling and admits exact closed-form solutions for problems with two and three objectives. We evaluate PCD within structured network compression settings, unstructured sparsity and low-rankness, and across a variety of synthetic experiments, showing Pareto dominance and better per-objective performance with secondary progress guarantees over existing methods, further exhibiting the interpretable trade-off that $τ$ provides.

Reinforcement Learning (RL) has enabled LLMs to excel in objective reasoning tasks such as mathematics and code generation. However, applying RL to open-ended tasks, such as creative writing, remains challenging because LLM-as-a-judge reward models often exhibit stylistic biases and positional inconsistencies, leading to unstable supervision. To address this, we propose OPERA (Objective Perplexity-based Reflective Alignment), which replaces unreliable external judges with intrinsic rewards derived from perplexity dynamics. Specifically, we derive an intrinsic reward signal from perplexity dynamics, quantifying uncertainty reduction at critical reflective states. During the cold-start phase, we introduce a data synthesis method that leverages carefully designed guiding words to generate diverse reasoning traces, along with perplexity-prioritized rollouts that utilize internal log-probabilities to identify logically consistent reasoning branches. This pipeline yields a large-scale dataset comprising 20,000 high-quality reasoning trajectories. Empirical evaluations consistently demonstrate the scalability and efficacy of our approach in alignment for open-ended tasks. Implementing OPERA on Qwen3-8B establishes a new state-of-the-art among open-source models, achieving parity with or surpassing proprietary models like Gemini2.5 and MiniMax-M2.5 in some open-ended tasks. The code is available at https://github.com/pangpang-xuan/OPERA.

Pre-trained language models struggle when applied to new domains, as full fine-tuning is computationally expensive and prone to catastrophic forgetting. This study addresses this challenge by presenting a novel parameter-efficient strategy for unsupervised domain adaptation that combines custom PEFT architectures with mixed-objective training. Our approach simultaneously optimizes classification performance on labeled source domain data and masked language modeling (MLM) on unlabeled target domain data, preserving target domain knowledge while adapting to source domain tasks. Our method employs a custom union of invertible adapters and Low-Rank Adaptation (LoRA) within a unified parameter-efficient framework. Through comprehensive evaluation on the Multi-Genre Natural Language Inference (MNLI) dataset across 20 domain shifts, our approach achieves significant improvements over existing methods: 1.41 percentage points over the current parameter-efficient state-of-the-art UDapter, 1.26 percentage points over the fully-tuned DANN baseline, and 0.86 percentage points over DSN, while utilizing only 7% of the model's trainable parameters. These results establish new benchmarks for parameter-efficient unsupervised domain adaptation and demonstrate that carefully designed PEFT combinations with concurrent optimization can outperform both existing parameter-efficient methods and traditional fully-tuned approaches.

We consider the problem of synthesizing control policies that enforce a concentration property on the state distributions of a stochastic system. We present a formalization of this problem in terms of synthesizing strategies for maintaining an entropy-based objective in Markov Decision Processes (MDPs). We first show that even relaxed versions of this problem are complexity-theoretically hard. We then present a sound and (conditionally) relatively complete method to verify and synthesize strategies for such entropy objectives. The main challenge is the non-linear nature of such objectives, and our approach addresses this by exploiting and combining ideas from convex duality and invariant synthesis. We also investigate the role of memory and randomization in ensuring entropy objectives. Finally, we implement our ideas to evaluate our approach empirically on a few illustrative benchmarks.

Aligning language models with human preferences often requires optimising multiple behavioural objectives. A practical approach is to apply these objectives sequentially using preference optimisation methods such as Direct Preference Optimisation (DPO), but it remains unclear whether later training uniformly degrades preferences learned earlier or whether the effect depends on the relationship between objectives. We study sequential DPO across four preference settings covering distributional conflict, multi-attribute interaction, strong safety signal, and compatible response-quality objectives. Using Llama-3.1-8B-Instruct with LoRA adapters, we evaluate all objectives after every stage with a fixed base-model reference. We find that sequential DPO does not produce a single forgetting pattern; preference change ranges from partial degradation to stability, pair-level redistribution, or positive transfer depending on objective relationship, signal strength, and training order. Pair-level analysis using length-normalised policy margins shows that aggregate metrics can mask heterogeneous changes across preference pairs, whereas quartile decomposition reveals that high-confidence pairs can either degrade or improve depending on the setting. Mechanistic diagnostics show that Stage~2 gradients and adapter updates are near-orthogonal to the previous objective across all settings, providing little evidence that direct gradient opposition is the primary driver. These findings suggest that future sequential alignment pipelines should account for objective compatibility and signal strength, rather than assuming that later objectives affect earlier preferences uniformly.

Foundation models (FMs) trained on large datasets and fine-tuned on downstream tasks have emerged as a powerful paradigm in AI for science. Industrial FMs are typically trained using self-supervision with masking due to the lack of labels. In many scientific domains, accurate simulations are plentiful and facilitate large, labeled datasets. This opens up new possibilities for pre-training. We present a systematic comparison of pre-training methods using the OmniLearned High Energy Physics FM framework. We test supervised classification, flow-matching generation, and self-supervised masked particle modeling. All models are pre-trained on the JetClass dataset and fine-tuned on two representative downstream tasks, top jet classification and JetNet conditional generation. Among other observations, for classification tasks, we find that pure classifier pre-training is optimal when downstream labels and model capacity are plentiful, but combining it with self-supervised masked particle modeling (MPM) is uniquely powerful in the low-finetuning label regime. Flow matching-based generative pre-training seems to provide little benefit for downstream classification, and interestingly, for downstream generation, we find that flow matching must be in the pre-training objective to see a significant finetuning advantage, hinting at the orthogonality of classification and generation tasks. That is, for a model to transfer to both generative and classification downstream tasks, it must be pre-trained on both. This study provides a template for controlled scaling analysis of pre-training objectives for foundation models in simulation-based sciences.

This position paper argues that contemporary AI paradigms are insufficient for supporting complex global goals and introduces Planet-Centered AI (PCAI) as a design philosophy and research agenda that reorients AI toward planetary-scale socio-ecological systems and their long-term trajectories. A planet-centered approach is grounded in systems thinking, treating Earth as an interconnected whole of which humans are part. We diagnose recurring limitations across AI frameworks, many of which remain human-centered, and show why these become especially consequential under current planetary conditions characterized by systemic risk, non-stationarity, and deep uncertainty. We then articulate how PCAI reshapes the AI lifecycle, from problem formulation and model design to evaluation and deployment, by emphasizing alignment with global agendas, developing system-aware AI foundations, trajectory-oriented evaluation, and monitorability. Finally, we advance a falsifiable claim: AI systems optimized without explicit consideration of systemic consequences are more likely to exacerbate systemic instability than to mitigate it.

Hindsight relabeling usually turns achieved future states into exact goals, which can overconstrain offline robot learning when task success depends only on a subset of the state. We propose Goal-Set Hindsight Relabeling (GS-HER), a predicate-level generalization of HER in which achieved states certify query-defined goal sets rather than singleton goal states. A binary query specifies which variables define success, making the goal predicate an inference-time input while leaving the underlying offline GCRL algorithm unchanged. Across OGBench tasks and five offline goal-conditioned learners, GS-HER improves performance when full-state goals are bottlenecked by nuisance dimensions and turns hindsight relabeling into a reusable goal interface: one checkpoint can answer multiple robot goal predicates without retraining.

Dialogue systems in e-commerce scenarios often need to satisfy multiple objectives: accurately reasoning over user profiles (e.g., eligibility, credit limit) to ensure correct decision-making and user state interpretation, while also generating natural and faithful responses. These goals are complementary but not identical. In this work, we propose MORE, an adaptive Multi-Objective REinforcement learning framework that jointly optimizes reasoning accuracy and linguistic naturalness. Our preliminary experiments show that directly mixing rewards with diverging optimization dynamics can cause oscillations and unstable learning. Thus, instead of optimizing a single mixed reward, we treat reasoning functions as constraints that guide policy optimization. At inference time, the system directly generates responses without explicit reasoning steps, while still benefiting from reasoning-enhanced scaffold and avoiding additional inference overhead. To better balance linguistic objectives during response generation, we introduce an adaptive multi-reward mechanism that aggregates signals such as fluency and naturalness and dynamically reweighs them via gradient feedback. We evaluate MORE on two real-world dialogue systems at ByteDance and the MultiWOZ 2.2 benchmark, where it consistently outperforms strong baselines. In 14-day online experiments on ByteDance production traffic, MORE improves overall and reached conversion by 16.53% and 30.09%, while increasing user satisfaction and reducing handoff rates. Notably, in a human-machine comparison, MORE recovers about 60% of the incremental conversion lift achieved by human agents.

Many current agentic systems and LLM pipelines correct mistakes by optimizing outcome reward. This addresses only the what of failure: when an outcome diverges from prediction, the why and when of the mismatch are not systematically logged, reviewed, or corrected, so the same error can recur episode after episode. We argue that this is a structural problem, not merely a model-capacity one. We propose long-horizon temporal regret as a first-class objective alongside outcome regret and epistemic regret over the working causal model. Temporal regret captures when failure persists: how long a miscalibrated causal model is tolerated before correction. Epistemic regret captures why failure persists: residual uncertainty or error in the working causal model. Together, the three regrets give a falsifiable account of what, why, and when a long-lived agent can fail. Modeling the agent as a stream of E episodes, we prove three conditional results under explicit causal-probing, persistence, and detectability assumptions. First, under observationally equivalent confounding, outcome-only learning cannot distinguish causal from spurious structure without an intervention channel, so temporal miscalibration can persist linearly even after outcome regret is driven to zero. Second, with a persistent causal log and budgeted probes, total probe complexity is logarithmic in the episode horizon, inducing O(log E) temporal regret. Third, under K detectable change-points, the rate extends to O(K log E). We instantiate Trivium and pre-register five falsifiable predictions. On CausalBench-Seq, Trivium follows the predicted logarithmic envelope while outcome-only baselines grow linearly. A pilot real-LLM stream provides preliminary external-validity evidence across one full E = 500 run and three E = 100 frontier-model pilots. Self-learning here means revising an external causal model, not retraining LLM weights.

Few-step distillation has emerged as a critical component in the development of advanced visual generative foundation models, substantially reducing inference overhead while enabling real-time generation and cost-efficient deployment across a broad range of practical scenarios. However, prior work has predominantly focused on advancing training objectives, while comparatively overlooking the training recipe, which has become increasingly critical in the era of large-scale foundation models. In this work, we systematically revisit the training recipe under the well-established distribution matching distillation (DMD) framework for both text-to-image generation and image editing, focusing on three key dimensions: training data composition, teacher guidance within DMD, and task mixture. Our empirical analysis reveals several non-obvious and counterintuitive phenomena, ultimately motivating the development of Qwen-Image-Flash. These findings highlight that effective few-step distillation depends not only on carefully designed objectives, but also on a principled training recipe.

Long-horizon reasoning requires a system to commit to medium-horizon intent without becoming rigid: re-plan too often and computation never coheres into multi-step structure; commit too long and the plan goes stale. We study this stability-adaptivity tradeoff in the latent reasoning setting, where multi-step computation occurs inside hidden state rather than externalized token traces. We extend the Hierarchical Reasoning Model (HRM) with a feudal-style manager-worker interface: a slow high-level module periodically emits a normalized directional subgoal that persists for P low-level steps, biasing the worker's hidden-state updates and supplying an intrinsic cosine alignment loss. On ARC and ConceptARC, we find that subgoal persistence -- not subgoal injection alone -- is the central knob: moderate periods P in [3, 6] consistently outperform both very frequent (P=1) and very long horizons, with a clear minimum LM loss at P=3 (1.544 vs. 1.674 at P=1, 1.640 baseline; replicated over 5 seeds at mean 1.595, std 0.045). The intrinsic alignment weight lambda shows a complementary narrow optimum (lambda approximately 0.05). A controlled ablation at past-sweet-spot lambda isolates learned directional structure -- not architectural capacity or auxiliary loss alone -- as the source of interference when the alignment signal exceeds its optimum. Together these findings implicate a design principle for compositional planning in latent reasoning systems: medium-horizon intent must be coherent across enough computational steps for compositional structure to form.

The Boolean Task Algebra (BTA) provides a principled framework for zero-shot task composition in reinforcement learning by equipping goal-reaching tasks with Boolean operations. We revisit its structural assumptions and formalize a collapse in the space of optimal extended Q-value functions: in deterministic MDPs, every such function is fully determined by the universal and empty tasks. This makes the logarithmic set of base tasks proposed in the original BTA formulation redundant. Building on this observation, we introduce a goal-set-based composition method that performs logical operations on goal sets and reconstructs composed value functions by selecting slices from the universal and empty value functions. This reduces learning costs for standard BTA and reduces composition time for both BTA and Skill Machines, while preserving policy performance. Experiments across tabular, visual, function-approximation, and continuous-control domains show that learning additional base tasks does not yield better performance. Finally, we study the stochastic setting and provide a counterexample showing that this collapse need not hold, that is, optimal composition may require accounting for exponentially many policies in the number of goals. Code is available at https://github.com/EduardoTerres/bta_paper.

Probabilistic models are typically trained using task-agnostic objectives like log-loss, which can lead to significant errors in downstream estimation. This disconnect is especially critical in Inverse Probability Weighting (IPW) for causal inference, where propensity score errors near $0$ and $1$ often lead to high bias and variance. We propose a principled framework for deriving task-specific strictly proper scoring rules by matching the local curvature of the downstream error metric. We apply this to the Average Treatment Effect (ATE) estimation, deriving a closed-form loss and its corresponding canonical probability mapping that can be readily integrated with any model like a neural network or a gradient boosting algorithm. Extensive evaluations on causal inference benchmarks demonstrate that our tailored objective consistently outperforms standard likelihood-based and covariate-balancing approaches.

Recommender systems have grown from content-organization tools into sophisticated systems that shape daily behavior. By controlling what we see, they shape what we perceive, raising concerns about filter bubbles, radicalization, polarization, and social inequality. Large language models (LLMs) enable more powerful personalization, intensifying these dynamics. Yet most recommenders are tuned for engagement or limited accuracy metrics, with little attention to broader social implications, e.g. how personalization reshapes exposure in socially consequential domains. We investigate whether LLM-assisted reranking, while improving personalization, inadvertently amplifies exposure to ideologically extreme or conspiratorial political content, a risk theorized but not empirically characterized in news recommendation. Using real news-consumption histories, we rerank YouTube's sidebar candidates through zero-shot, instruction-based prompting. We compare a baseline prompt with a constrained variant that preserves topical relevance and broadens ideological exposure while reducing conspiratorial or extreme content. Without constraints, reranking strengthened personalization but increased exposure to conspiratorial and extremist material for users whose histories contained such content. Lightweight prompt-level regularization reduced promotion of extreme content and increased ideological diversity, with modest relevance loss. Synthetic experiments suggest that LLMs rerank via statistical regularities in language rather than semantic understanding of ideology, clarifying why naive prompts amplify these patterns and why regularization can reshape them. Together, our results highlight the power of LLMs to operationalize contextual nuance in high-stakes recommendation, and the need to evaluate LLM-assisted personalization beyond accuracy and treat prompt design as a value-laden rather than neutral default.

Reinforcement Learning (RL) has long served as a model for goal-directed animal behavior in neuroscience. Modern deep RL has shown remarkable success across many domains, further strengthening this connection. The ability to learn abstract representations of high-dimensional state spaces underlies much of this success. However, theoretical understanding of these learned representations remains limited, hindering direct comparisons between models and animal learning. We address this gap by analyzing deep RL representations through the lens of MDP reduction theory. Investigating canonical RL algorithms in a navigation task, we find that even when performance is comparable, the value-based method (DQN) learns representations that are invariant to MDP homomorphism symmetries, while the policy-gradient method (PPO) learns representations invariant to action symmetries. These differences emerge consistently across domains, have downstream consequences for transfer learning, and appear in LLMs in a prompt-dependent manner. Our findings provide a principled approach to comparing learned representations across RL algorithms, with demonstrated practical implications and possible insights for neural coding in the brain.

Multi-agent systems are increasingly deployed to support various tasks where agents interact to achieve individual and collective objectives. Although these systems can enhance task performance and decision-making, fairness preservation through bias reduction remains challenging. This study examines how agent-level biases shift and impact system-wide fairness. We use prompts to expose individual agents to group-favoring bias, then assess downstream impacts at the system level. To quantify the impact, we propose Favor Bias Strength (FBS), a zero-centered metric that decomposes bias alteration between favored-group uplift and disfavored-group suppression. Using multiple agent designs, benchmarks, and up-to-date large language models, we show that agents endowed with bias can substantially affect system-wide fairness. Interestingly, when agents are exposed to bias uniformly, the system-wide bias elevates, even exceeding the additive sum of the individual agents' biases. The empirical evidence underscores the criticality of fairness in multi-agent systems, which warrants further analyses and empirical tests.

Large language models (LLMs) learn undesirable properties during pretraining, including dangerous knowledge and toxic text generation. Just as post-training uses different objectives to shape different behaviors, we argue that unlearning methods should be designed for the language function at issue. To study this, we consider two mechanistically distinct unlearning goals, dangerous-knowledge unlearning and toxicity unlearning. For dangerous knowledge, we introduce a cosine-based, meta-learned variant of RMU. For toxicity, we propose a multi-layer objective based on layer-specific probe directions. Across four open-source 7-8B models, our methods achieve strong results, based on distinct training objectives for the two types of unlearning. Overall, our results suggest that unlearning should be studied as a family of problems, analogous to the multiple types of LLM post-training.