Large Language

Large Language Models (LLMs) are increasingly applied to legal and criminal justice tasks, yet existing work focuses almost exclusively on post-arrest scenarios where the suspect's identity is already known, leaving the critical pre-arrest challenge of inferring suspect characteristics from incomplete evidence largely unexplored. To fill this gap, we introduce the Profiling, Investigation, and Judgment (PIJ), comprising 2,500 real homicide cases from five countries. PIJ evaluates LLMs across three tasks that span the entire criminal investigation pipeline: criminal profiling, which requires abductive reasoning to infer suspect attributes from fragmentary scene evidence, crime process reconstruction, which tests structured information extraction, and sentence prediction, which demands legal deductive reasoning. We evaluate 9 powerful LLMs and find that performance degrades systematically as tasks shift from explicit fact extraction to implicit reasoning over unknown suspect profiles. Categories requiring inferential reasoning, such as motivation and victim-offender relationships, remain the primary bottlenecks. Further analysis reveals substantial gaps between LLMs and human experts, along with pervasive biases in gender, age, and motive attribution. Our findings indicate that pre-arrest inference from incomplete evidence remains an open challenge.

Modern applications generate massive volumes of raw telemetry data, but translating those noisy, heterogeneous event streams into actionable business insights remains a fundamental challenge. Data engineers and analysts expend substantial effort reconciling semantic discrepancies, hand-crafting parsing logics, and maintaining fragile mappings between raw data and business KPIs. In this paper, we present an end-to-end framework that fully automates the construction of a business semantic layer from application raw logs. Our approach introduces a two-stage semantic abstraction: first, high-level business features are identified via LLM inference augmented with domain-specific industry knowledge; second, fine-grained business nodes are derived through a structured pipeline comprising data refinement, hybrid retrieval, multi-stage filtering, semantic clustering, and canonical naming. Evaluation on production-scale telemetry demonstrates that our system improves human-assessed semantic quality from 50 to 80+ on a 100-point scale, reduces maintenance effort by 80%, filters out 74% of noise, and achieves 0.87 Cohen's kappa via an integrated LLM-as-Judge evaluation, enabling continuous, scalable quality assurance. Overall, our work distinguishes itself from prior work by addressing the novel problem of business semantic layer induction from raw telemetry, operating without labeled training data or manual rule engineering.

We present a robust benchmark for measuring cultural value orientation in large language models on the two Inglehart-Welzel axes over several domains and roles (hence GYROval - Gridded Yielding of Robust value Orientation), together with the results of administering it to twenty models. Items are binary contrastive scenarios in the sense introduced by CDEval: both options are legitimate courses of action, neither is correct, there is no answer key, and a model's score on an axis is the proportion of its responses falling on the counted pole. Eleven of the twenty models were additionally administered a paired Russian translation of the identical items and a second sampling temperature. The instrument is publicly released in both languages. Stability was assessed by treating the vignette as the unit of analysis, ranking the models within the levels of each perturbation factor, and summarising the agreement between levels by tie-corrected Kendall's \emph{W} against an empirical permutation null.

Payment operations are a critical financial infrastructure, but the value of large language models in this domain remains unclear because payment rules change quickly, evidence is fragmented, and decisions depend on transaction state, participant role, region, and payment rail. Existing benchmarks do not isolate whether failures come from missing payment-rule knowledge, poor use of supplied evidence, or brittleness under imperfect harness inputs. We introduce BENCHCOMPASS, a payment-domain benchmark whose construction pipeline builds scenario-grounded tasks from typed evidence packs, applies LLM-based quality checks, creates task-input attack variants, and reserves final item admission for domain experts. The release contains an expert-reviewed Pro benchmark covering payment knowledge, context-grounded scenario reasoning, and Attacked Open robustness, plus a lower-assurance Normal pool for inspection and future curation. Across 16 model variants, BENCHCOMPASS shows qualitatively different failure modes: missing parametric payment knowledge, incomplete reasoning over supplied rules, and failure to reject plausible but invalid workflows. The benchmark remains unsaturated: the best frontier model reaches 89.6% on Open Context-Grounded Reasoning and 81.7% under attacked inputs, while a representative 32B open-weight model reaches 69.8% and 42.6%. Benchmark data and code are available at https://github.com/ant-intl/BenchCompass.

AI-driven penetration testing has been demonstrated with premium frontier models such as GPT-4, but the per-engagement token cost makes continuous, automated testing unaffordable for the smaller organisations that need it most. This paper presents PentestChain, a ten-phase automated penetration testing framework that couples a curated, deterministic exploit map with a cost-aware AI cascade-a local Ollama model (qwen2.5-7b) first, then free-tier OpenRouter and Cerebras, with a rule-based fallback that always produces output-and exposes the full pipeline through a Model Context Protocol (MCP) server with eleven tools. We make three contributions. First, we treat US-dollar cost per engagement as a measured, first-class evaluation metric and show that a 7B-parameter local model, kept off the critical path by a deterministic backbone, sustains end-to-end operation at zero measured paid-API cost. Second, we analyse the attack surface that an MCP-exposed offensive engine introduces, grounding a four-position threat model in the 2025 MCP incident record (the CVE-2025-6514 remote-code-execution flaw in mcp-remote, the postmark-mcp supply-chain backdoor, and the tool-poisoning-rug-pull-line-jumping class), and contribute four mitigations. Third, we specify a reproducible, containerised evalua-tion protocol aligned with the standardised testbeds now expected at top-tier venues-AutoPenBench, a Cybench subset, and the PentestGPT 182-sub-task benchmark-with multi-trial statistics (more than 10 trials per configuration, pass-at-k, non-parametric significance tests and effect sizes) and direct, same testbed reproduction of the PentestGPT and PentestAgent baselines rather than citation of their published numbers. On the legacy targets measured to date, the framework detected 26 services, enriched 34 CVEs, produced

This study empirically evaluates whether cost-efficient Large Language Models (LLMs) can be trusted to generate enterprise code to a written specification. Three models (Gemini Flash 3, GPT-5.4 mini and Claude Haiku 4.5) were asked to solve 992 algorithmic problems as Java Spring Boot service methods conforming to a mandated signature and data-transfer-object specification, crossing four model and agentic coding tool combinations with two prompt variants to yield eight configurations, with iteration forbidden and hardcoded answers explicitly prohibited. Eight problem statements were withheld to probe how models respond to missing input. The 7,593 resulting methods were classified by an eight-class outcome taxonomy describing what each does about producing an answer, then deployed and executed, giving 7,936 measured requests joined to that classification. Structural conformance approached ceiling, yet 38.4% of methods do not compute the value they returned and only 12.9% of returned answers were correct. Conditioning on outcome class shows that response reliability and correctness are inversely related, whereas genuinely computing methods answered least often and were correct 19.3%. Limitations include single generation runs per configuration, partial harness coverage, single-pass timing, syntactic classification, and probable corpus contamination.

This paper addresses the challenge of making complex healthcare information more accessible through automated Plain Language Adaptation (PLA). PLA aims to simplify technical medical language, bridging a critical gap between the complexity of healthcare texts and patients' reading comprehension. Recent advances in Large Language Models (LLMs), such as GPT and BART, have opened new possibilities for PLA, especially in zero-shot and few-shot learning contexts where task-specific data is limited. In this work, we leverage the capabilities of LLMs such as GPT-4o-mini, Gemini-1.5-pro, and LLaMA for text simplification. Additionally, we incorporate Mixture-of-Agents (MoA) techniques to enhance adaptability and robustness in PLA tasks. Key contributions include a comparative analysis of prompting strategies, finetuning with QLoRA on different LLMs, and the integration of MoA technique. Our findings demonstrate the effectiveness of LLM-driven PLA, showcasing its potential in making healthcare information more comprehensible while preserving essential content.

This study proposes a Large Language Model-Driven Differential Evolution (LLMDE) algorithm to reduce the reliance on handcrafted hyperparameter design. The proposed algorithm leverages a prompt engineering strategy, allowing large language models (LLMs) to dynamically select mutation strategies and configure control parameters guided by optimization feedback, thus enhancing the performance of the DE algorithm. We evaluate the performance of LLMDE on the CEC2022 benchmark suite, comparing it with standard DE and representative metaheuristics. Furthermore, we employ factor analysis and K-means clustering for stock selection, and then apply LLMDE to solve the Conditional Value at Risk (CVaR) portfolio optimization problem using the selected stocks, subject to budget and minimum expected return constraints. Experimental results demonstrate that LLMDE achieves competitive performance on the benchmark suite while continuously generating high-quality solutions for complex constrained optimization tasks. These outcomes successfully demonstrate the viability of embedding LLMs within metaheuristics, paving a promising path toward the design of advanced LLM-assisted optimization techniques.

Ternary Large Language Models (LLM) store every weight as one of three symbols $\{-1,0,+1\}$, so the cost of a ternary model is conventionally referenced to the information-theoretic $\log_2 3 \approx 1.585$ bits per weight. The prevailing deployment format packs five ternary weights into one byte (five-trit packing), and due to the power-of-two group sizes used in practice this rounds up to $1.625$ bits per weight. This effective storage bit-width treats the three symbols $\{-1,0,+1\}$ as equiprobable. We measure the actual symbol distribution of 29 ternary LLM models and find that zeros account for up to $51.5\%$ of all weights. Motivated by this finding, we introduce BITCOS, a simple distribution-adaptive layout comprised of a dense presence bitmap plus a compacted sign vector, and costs $2 - z$ bits per weight element given a zero density $z$ in the model's weights. BITCOS stores weights more compactly than the five-trit packing in 26 of the 29 tested models, and reaches $1.485$ bits per weight on the sparsest of them. BITCOS is amenable to efficient unpacking on modern processors and GPUs, and we present optimized unpacking sequences for AVX-512, AVX2 and Intel Xe2 GPUs. Measured against production state-of-the-art ternary matrix-vector multiplication kernels, at the zero densities real-world ternary models exhibit, the realized gain with our proposed layout is up to $1.28\times$. Finally, we illustrate end-to-end LLM inference results on 5 different platforms (client and server CPUs, integrated and discrete Xe2 GPUs) where decode throughput improves by up to $1.18\times$ on CPUs and $1.27\times$ on GPUs.

Language model safety is typically evaluated one interaction at a time. We show that a weaker, unaligned model can split a harmful task into benign-looking subproblems, consult a stronger aligned model independently on each, and combine the answers locally. We call this attack capability laundering. Unlike a jailbreak, no single response is a harmful task. We measure consultation-aided uplift using tasks that a raw frontier model solves, the aligned frontier refuses, and the unassisted orchestrator fails. We evaluate GPT-5.5, Claude Opus 4.8, and Grok-4.3 as consultants to four local orchestrators on CyBench, BountyBench, and harmful CBRN requests. On CyBench, Gemma-4-31B recovers 8/14 candidates with GPT-5.5 and 7/9 with Opus, compared with 2/21 and 4/15 for Gemma-4-12B. On BountyBench, Gemma-4-31B recovers 3/9 and 2/3 candidates, while Muse-Glimmer-30B recovers none of 22 and 13. For CBRN, we measure uplift across eight steps of a hypothetical bioweapon attack chain and find that consultation raises Gemma-4-31B's mean rubric score from 62.3 to 83.1 on a 100-point rubric scale. These results expose a gap in current defenses: refusing a harmful task does not prevent frontier capabilities from being transferred and composed across many individually permitted interactions.

We introduce the Universe of Universes (UoU) framework, which treats the full ecosystem of major large language models (LLMs) as a structured retrieval corpus and proposes a compositional Automated Reasoning (AR) and Machine Learning (ML) architecture for cross-model retrieval-augmented generation. The central contribution is the formal characterization of the Benefit Yield Function (BYF), the marginal performance gain per additional model added to an ensemble, and the identification of the implosion threshold θ*: the ensemble size at which BYF crosses zero and aggregate performance begins to degrade. Existing LLM ensemble and mixture-of-agents systems treat models as responders and aggregate outputs, but do not study performance as a function of ensemble size N across the full model universe. Benchmark research confirms performance plateaus at the individual model level; model collapse literature establishes that iterative training on AI-generated outputs degrades individual model distributions. Neither body of work formalizes the ensemble-level implosion threshold, models Epistemic Hereditary Drift (EHD) at the ecosystem level, or treats AI manufacturing velocity as a co-variable of θ*. The framework has direct implications for DoD multi-model AI acquisition policy and the emerging science of testing AI-enabled systems.

Environmental, Social, and Governance (ESG) reporting is critical for corporate accountability, with Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) offering strong potential to automate KPI extraction. However, open-source LLM performance in domain-specific ESG tasks remains insufficiently understood. This paper evaluates open-source LLMs in ESG contexts using a structured framework and evaluation resource based on 498 real-world ESG reports from EU-listed companies (2010-2024). We evaluate seven open-source models (2B to 30B parameters) -- glm-4.7-flash, nemotron-3-nano:4b, qwen3:4b-instruct, gemma3:4b, gemma4:e4b, gemma4:e2b, and ministral-3:8b -- using 100 persona-based synthetic QA pairs covering ESG information needs. System performance is assessed via RAGAS metrics, including contextual recall, precision, relevance, faithfulness, answer relevancy, and factual correctness. Results show notable performance variations across architectures. Retrieval performance is strong across models (context recall around 0.58-0.61, context precision around 0.78-0.81, context relevance 0.965-0.985). Generation diverges most on faithfulness (0.607-0.822) and least on answer relevancy (0.760-0.881): glm-4.7-flash leads in faithfulness (0.822), qwen3 in factual correctness (0.449), and ministral-3 in answer relevancy (0.881). Low overall factual correctness (0.387-0.449) highlights the need for domain-specific fine-tuning. This work provides data-driven guidance for deploying open-source models in ESG reporting.

A compact controller can coordinate stronger experts by selecting whom to consult, formulating requests, and integrating their responses. We study whether learning from both the controller's decisions and the experts' reasoning and code improves its generation after expert removal. We introduce \textsc{Rivet} for \emph{collaboration internalization}: expert-augmented reinforcement learning applies a shared outcome signal to controller decisions and returned expert spans, and verified trajectory internalization consolidates complete successful interactions through format-aware supervised training. The deployed controller generates reasoning, code, and interaction structure with local Python execution and no external LLM. Across seven competition-mathematics benchmarks, RIVET-1.7B and RIVET-4B achieve average accuracies of $28.25\%$ and $44.16\%$; Stage~II improves RIVET-4B's accuracy after expert removal by $6.49$ points, and GPQA-Diamond results provide evidence of generalization to scientific reasoning. Ablations show gains from ordinary trajectory supervision and additional format weighting, supporting the effectiveness of training on the content and structure of verified collaborations.

Traditional large language models (LLMs) are scoped to individual user sessions, limiting their knowledge to a single conversation and preventing them from learning user preferences that evolve over time. Existing agentic memory systems address this limitation but generally operate at the individual-user level, restricting the public knowledge that could be shared across users to improve downstream responses. We introduce AIM (Agentic Interoperable Memory), a unified, privacy-aware memory framework that enables multi-agent, multi-user LLM systems to persistently manage private and shared memory. AIM dynamically classifies information as private, scoped to one user and inaccessible to others, or public, accessible to all users. It enforces index-level access controls so that private memories are retrievable only by their owner, protecting sensitive data while allowing beneficial shared knowledge to improve coordination and consistency. We also introduce MUMBench (Multi-User Memory Benchmark), a dataset of multi-user interactions containing private and shareable information across four domains. To our knowledge, MUMBench is the first public dataset designed to evaluate multiple memory operations, including retrieval, creation, update, and deletion, in a multi-user environment. Across three independent runs on MUMBench, AIM achieves 96.0% visibility classification accuracy, 58.8% strict operation accuracy, and 70.5% state-aware operation accuracy.

Post-training with reinforcement learning (RL) is a critical phase in the development of code-generating large language models (LLMs), as it ensures adherence to instructions and the production of functionally correct code. This process typically requires computationally intensive code sample generation from Transformer-based LLMs and substantial GPU-CPU communication for sequence verification. To address these computational challenges, this work examines whether RL-based post-training can be performed entirely offline by leveraging existing datasets rather than generating new samples. The findings indicate that, with only a few hours of training, zero-shot code generation performance of LLMs can be substantially improved without online sampling. Additionally, offline RL produces performance gains across models ranging from 0.5B to 7B parameters, although the extent of improvement varies among model families.

Background. Large language models are increasingly used to help write biomedical text but may fabricate references to nonexistent work. How often large language models do so is not well characterized. Methods. We prompted 26 language models from eight developers (2023 to 2026) to supply a missing reference for each of 69 biomedical passages across ten domains. References were classified as verifiable (real paper with a resolving identifier), partial matches (real paper without a resolving identifier), fabricated (no matching indexed paper), or declined (the model refused to supply a reference). A reference was considered correct in every evaluated bibliographic field only when it was verifiable and its journal, year, and listed authors matched those of the cited paper. Results. Fabrication ranged from 10.2% (Claude Opus 4.8, which declined 52.1% of prompts) to 98.4% (Ministral 3B, which produced no verifiable reference). Claude Opus 4.6 and Claude Sonnet 4.5 produced similar proportions of verifiable references (77.6% and 76.6%) but named authors correctly in 78.7% and 28.7% of author-evaluable verifiable references, respectively, and were correct in every evaluated field in 54.6% and 19.9% of responses. GPT-5.5 was correct in every field in 48.1%. Across all models, 55.4% of responses were fabricated and 14.9% were correct in every field. Among the five tested models first released in 2026, the corresponding proportions were 35.3% and 31.8%, respectively. Conclusions. Fabrication remained common, and no model was correct in every evaluated bibliographic field in more than 54.6% of responses. Models that identify real papers may still misstate their metadata, so references produced with model assistance require verification before use.

Allowing large language models (LLMs) to retrieve information from a set of trusted documents can increase reliability and reduce hallucination. However, recent work has demonstrated that retrieval-augmented generation (RAG) can have unintended side effects on the overall safety of the generated responses, when prompted for harmful or dangerous content. A clearer understanding of the mechanisms leading to this result is needed, as increasing numbers of end users turn to RAG to incorporate corporate documents and knowledge bases into LLM-based systems. We introduce RAG-Safety-Bench, a benchmark to measure the safety impact of RAG on LLM models. By removing the confounding effect of retriever quality, and cleanly separating the problem into four conditions -- non-RAG, RAG with an oracle document containing the answer to the harmful request, RAG with documents related to the harmful request but without the specific answer, and RAG with random, safe documents -- the benchmark isolates the impacts of different factors in the observed safety degradation. We report results across five open-source LLMs, showing an inverse relationship between benign and unsafe capability, strong evidence that baseline safety guardrails do not lead to downstream safety guarantees in the RAG case, and model-specific support for previous findings that even benign documents can lead to unsafe generation in retrieval-enabled systems.

We placed 100 memory-equipped large language model (LLM) agents in charge of a closed, money-conserving spatial economy on real Pokhara Lakeside geography (earning wages, running businesses, setting prices) and ran this multi-agent simulation for up to 26 simulated weeks, well past the 1-2 weeks typical of agent-society studies. Across 91 validated runs (2.44M agent decisions, 21.5B tokens), the money stops moving, in a specific and measurable way. A 12x tourist demand shock raises business revenue 4.62x ($p<0.001$), which we decompose exactly into a 1.50x extensive margin (more businesses trading) and a 3.07x intensive margin (more revenue each). Monetary transmission stops there. Wages move 1.03x ($p=0.42$); 0.3% of 3,981 menu items are ever repriced ($p=0.47$). A randomized cash transfer (NPR 5,000 to 20 of 100 agents) shows the same pattern from the opposite direction: 96.7% is still held 311 pulses later, marginal propensity to consume 3-4% by two independent measures, indistinguishable from zero. The wealth distribution is consequently near-frozen at the horizon this literature uses ($\rho=0.964$ over 2 simulated weeks), but not frozen. $\rho$ falls to 0.832 at 12 weeks and 0.752 at 26, a horizon-dependence no short study can see. Matched ablations show which knob actually matters. Swapping the backing LLM moves every outcome we measure ($p=0.0039$); deleting agents' memory moves none of them detectably. A purely social tool fails 94-97% of the time across two model families, compared with ~96% success on economic tools, with no measurable shift away from it. Every headline number is verified twice, by a live validator and by an offline recomputation that reconciles each agent's wealth against its own signed transaction history, and we release the full run corpus for reanalysis.

Datacenter GPU power is the binding constraint on LLM serving capacity, and production serving has shifted to prefill/decode (PD) disaggregation. Deploying NVIDIA's Max-Q inference profile on a disaggregated B200 system, we found its realized gain modest (+8.6% tokens/J), model-dependent, and carrying a mean end-to-end latency cost (+5.2%) that throughput-only evaluation does not surface; the profile also applies one setting to prefill and decode GPUs that operate in opposite hardware regimes. We hypothesize that the optimal power setting is a property of the deployed (model, quantization, engine, hardware) combination rather than of the GPU class, that each lane warrants its own profile, and that converting SLO headroom into energy safely requires latency-gated calibration under a runtime SLO guard rather than a fixed recipe. We present a phase-decoupled, model-calibrated controller: the prefill lane runs under an SM-clock window whose floor is a latency guarantee by construction, and the decode lane under a power cap placed by automatic calibration just above a measured throughput/latency cliff. Because a disaggregated decode lane draws flat, memory-bound power, the cap binds continuously, the reactive-overshoot weakness that led POLCA to reject capping is absent, and the GPU's own power manager retains throughput under the cap. On an 8x B200 node serving Qwen3-Coder-480B (FP8) under agentic load, our balanced mode delivers +20.4% tokens/J at +3.5% mean e2e versus +8.6% at +5.2% for Max-Q, a Pareto improvement on both axes. On Qwen3-235B-A22B (NVFP4) every operating mode meets the ITL-p99 SLO in every repetition; both vendor profiles miss it. A decode-actuator A/B shows the calibrated cap beats static clock locks, and a three-day sustained run saves 32.3% of a lane pair's electricity. Both models are MoE; a dense model recovers roughly 5x less, so we scope our claims to MoE serving.

Recent large language models can emit task-progress signals that agent frameworks use to decide whether a task should continue or stop, yet whether a model can reliably report its task progress at every stage of a task, and where and how its reports fail, has not been studied systematically. We evaluate this ability on the public benchmark $τ^2$-bench and on StageIF, a controlled testbed in which reporting checkpoints are placed across the task's lifecycle. Both settings require reports at multiple task stages. We find that reporting reliability depends on the stage a task has reached, and that almost every deployed model we test is reliable at some stages and unreliable at others. Where reporting breaks down is not the same everywhere. Most deployed models lose accuracy once work is under way and recover once the task is done. The newest generation closes that mid-task drop and instead grows conservative at the finish line. Our study exposes a capability gap in task-progress reporting and provides an evaluation protocol that spans the whole course of task execution for this ability on which agent operation depends. The findings indicate that agent frameworks should not control task flow on the strength of the model's state reports alone.

As AI-code assistant tools become widespread, automatic assessment of the correctness of generated code becomes a significant challenge. Code LLMs are prone to hallucinations, which may lead to code that does not solve the required problem, or even to code with severe security vulnerabilities. In this paper, we introduce CodeTD -- the first approach to pre-execution assessment of code correctness based on topological data analysis (TDA) of Code LLMs' attention maps. Our method quantifies prompt-generation mismatch using topological patterns of attention maps. We carry out experiments with common benchmarks (HumanEval, MBPP, BigCodeBench, MultiPL-E), 5 programming languages and 10 Code LLMs of size up to 34B parameters. The experimental results show that the proposed method outperforms recent baselines. Moreover, CodeTD is transferable between coding benchmarks.

Financial large language models are increasingly deployed for summarization of reports and disclosures, where numerical hallucination poses significant practical risks. While prior work often attributes such hallucination to insufficient numerical reasoning, this assumption has not been systematically tested under controlled fine-tuning settings. In this paper, we conduct a cost-effective, controlled study of numerical hallucination in financial summarization across three model variants: a base instruction-tuned model, a domain language-adapted model (FT-A), and a numeracy-enhanced domain model (FT-A+B+C). We introduce a three-level detectability taxonomy distinguishing between overt hallucination (currency-denominated fabrication), covert-explicit hallucination (professional-convention numbers), and covert-implicit hallucination (ungrounded quantitative claims). Our results reveal that domain fine-tuning substantially degrades numerical restraint at all detectability levels. While the Base model maintains near-zero hallucination rates (5.4%), FT-A exhibits 82.5% overt hallucination and FT-A+B+C reaches 98%. Contrary to intuition, numeracy supervision amplifies rather than mitigates hallucination across all levels. We identify template injection---the insertion of memorized canonical values regardless of input content---as a primary hallucination mechanism in fine-tuned models. These findings demonstrate that numerical hallucination in financial summarization is driven by the degradation of numerical restraint through domain adaptation, not by insufficient numerical reasoning. We recommend that evaluation protocols assess hallucination across all detectability levels and that deployment practices include explicit mechanisms for grounding-aware generation or abstention.

On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher. Existing work has mainly studied its algorithmic behavior, leaving the role of training data unclear. We examine this role at the data-minimal limit by training on a single query. One-shot OPD keeps improving for hundreds of steps and recovers most of full-data OPD's gain across task domains and model families. We explain this result through the states visited during training and the rate at which the student aligns with the teacher. We measure \emph{state coverage}, the fraction of the states full-data OPD visits that a query set's rollouts reach. A single query already reaches $71.5\%$, most of it within the first 100 steps. Adding semantically distinct queries raises coverage and validation accuracy together, until 16 queries reach $98.9\%$ and match full-data training. Yet alignment slows at a similar pace whether OPD trains on one query or the whole dataset, and even a fixed set of states takes hundreds of steps to absorb. OPD is therefore data-overfed but algorithm-starved. Its rollouts quickly expose broad supervision, while the student absorbs that supervision increasingly slowly. The state-coverage result extends to multi-teacher OPD, where 16 semantically diverse queries per domain match full-data MOPD. As a further stress test, content-light templates and off-domain WildChat queries also approach the real-query baseline. Task content and induced state coverage can therefore come apart. We hope these findings direct future work toward the step efficiency of OPD, and prompt a re-examination of the data and the mechanisms behind its recent successes in frontier post-training.

Large language models (LLMs) have become ubiquitous tools for code generation and editing. However, development teams often use multiple LLM assistants. Different developers may prefer different models, and individual developers may switch between models across different coding sessions. Because of this, the edits any one model makes are frequently applied to foreign code originally generated by another model. These LLMs are often trained on different datasets, and as a result have different stylistic preferences. Do LLMs behave differently when they edit foreign code originally written by a different LLM with a different coding style? We find that models tend to make more, and often excessive, edits on foreign code. We introduce CROCODIL (Cross-model Code Editing with LLMs), a post-training framework for reducing excessive edits while preserving functional correctness. CROCODIL's similarity reward penalizes large changes, while its execution reward scores build and test success. We use the product of these two rewards to encourage the policy to decrease the edit size without decreasing the edit task success rate. CROCODIL is available at https://github.com/EngineeringSoftware/Crocodil.

Long-output reasoning has made the key--value (KV) cache a critical memory bottleneck for efficient LLM serving. Existing KV compression methods usually rely on a predefined per-request budget and adjust only which KV states are retained, leaving the total capacity fixed throughout decoding. However, reasoning workloads exhibit substantial demand variation: different requests require different KV capacities, and the attention demand of an individual request evolves during generation. We introduce \textbf{GrowPage}, an on-demand KV budgeting framework that treats KV capacity as a runtime resource. GrowPage maintains lightweight dual-timescale query summaries to capture recent and long-term attention behaviors, and uses their relative attention working sets to estimate demand evolution. At each capacity boundary, GrowPage either compresses KV states within the current allocation or acquires an additional physical page when broader demand emerges. By integrating with PagedAttention's page-level memory abstraction, GrowPage preserves continuous batching and prefix caching. Experiments on reasoning benchmarks across multiple models show that GrowPage achieves a superior performance--throughput trade-off over existing approaches.

Idiomatic expressions are an integral part of natural language, reflecting cultural nuances and posing unique challenges for computational models, particularly in low-resource languages. In this paper, we present the first large-scale benchmark dataset of Bangla idioms, complemented by a synthetic multiple-choice question (MCQ) dataset for idiom meaning identification. We conduct a comprehensive evaluation of recent large language models (LLMs) across three idiom-related tasks: paraphrasing, idiom span detection, and meaning identification, leveraging zero-shot and few-shot prompting strategies. Our results reveal substantial variability in model performance, with no single LLM consistently outperforming others across all tasks. Notably, Phi-4-mini-instruct excels in paraphrasing, Kimi-K2-32b-instruct in span detection, and Gemini-2.5-flash in meaning identification. We believe that our datasets and analyses will provide valuable resources to guide future research in improving LLM comprehension of idiomatic expressions, particularly in Bangla and other low-resource languages.

Scientific citations carry rhetorical intent. Scholars may cite prior work positively (supporting), negatively (contrasting), or neutrally (mentioning). As large language models (LLMs) increasingly assist scientific writing, whether they reproduce citations with the same rhetorical intent as humans remains unclear. We introduce a masked-citation task to compare human and LLM-generated citation behavior. For each citation context, an LLM generates a replacement citation sentence, producing a counterfactual corpus directly comparable to human citation. We analyze what, whom, and how models cite, using an LLM-as-a-judge to classify citation intent and a 20-million-edge coauthorship network to measure social distance between cited authors. Across six popular LLMs and 1,746 top NLP conference papers (63k+ contexts, 132k+ citations), three patterns emerge: (1) Compared with human citation, LLMs cite significantly less critically; (2) LLMs over-cite popular and older papers, a tendency amplified for contrasting citations where human writing more often draws on recent, niche work; (3) Whereas humans often cite within their close social network, especially for supporting citations, LLMs tend to draw on more socially distant authors. Together, these differences are double-edged: LLM citation reaches beyond a scholar's close collaborators while being less critical and amplifying visibility bias, reshaping the rhetoric and reach of scientific citation.

Current Large Language Models (LLMs) are primarily optimized for written text, often producing outputs that are grammatically correct and helpful yet poorly suited for spoken delivery via Text-to-Speech (TTS). In this work, we study how to make LLMs natively generate TTS-friendly text, which we frame as a preference alignment problem: instead of relying on downstream rewriting modules, we directly align LLMs to generate text optimized for spoken delivery. We introduce two preference datasets spanning different target domains, CORA and Recipe, which contain paired TTS-friendly and TTS-unfriendly responses. We further propose an evaluation suite combining a pattern-based heuristic metric, a TTS$\to$ASR evaluation pipeline, and a MUSHRA listening study with human judges. Our experiments compare the recently proposed Feature-aware Sampling and Tuning (FaST) framework -- leveraging interpretable features instead of a black-box reward model -- against an array of alignment baselines on the TTS-friendly generation task. Notably, we found that FaST achieves the best overall tradeoff between TTS-friendliness and helpfulness across various settings. We also identified a strong correlation between our different metrics, highlighting the ability to reliably assess TTS-friendliness via an efficient heuristic.

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.

Weight-only post-training quantization (PTQ) can alleviate the computational burden of serving large language models (LLMs) at scale. However, existing PTQ methods often fail to generalize across models and suffer severe accuracy loss below 2 bits. Many leverage unstructured sparsity to mitigate this loss, but at the cost of regularity and GPU-friendly execution. We present QTEA, a sub-2-bit PTQ framework that quantizes weights into ternary values and uses salient weights as residual error compensators. To maintain hardware efficiency, residuals are assigned to selected columns with semi-structured $1:4$ sparsity within the salient columns. We further add column-wise rescale refinement to GPTQ-style column-by-column quantization, alternately updating per-column scales and ternary assignments to reduce reconstruction error. We also identify order-dependent error propagation in GPTQ and introduce error decay to attenuate late-stage error accumulation. On Qwen3-14B, QTEA compresses all weights to an effective 1.7 bits per weight while improving average accuracy over the strongest ternary PTQ baseline by 16.7%. It also achieves 1.40$\times$ and 2.61$\times$ lower perplexity on WikiText and C4 respectively. This trend holds on Llama3-8B, where QTEA obtains a 6.6% accuracy gain and 1.34$\times$ / 1.95$\times$ lower perplexity on the same datasets. Finally, we develop a lookup-table based kernel that achieves 7.2$\times$ faster per-token generation over an FP16 baseline. Code is available at https://github.com/Intelligent-Microsystems-Lab/QTEA.

Reinforcement learning (RL), particularly RL with Verifiable Rewards (RLVR), has recently emerged as a central paradigm for enhancing large language models' (LLMs) reasoning abilities, demonstrating remarkable effectiveness across reasoning tasks. Recent studies suggest that high-entropy tokens play an exceptionally important role in model training, since training with only the highest 20% entropy tokens yields significant performance gains. However, why such high-entropy tokens are beneficial remains insufficiently understood. In this work, we find that although high-entropy tokens within one answer tend to correlate with large gradient magnitude, entropy alone fails to consistently reflect token importance across different answers, considering the variations in the answer-level reward signals. Based on this observation, we introduce the Gradient Magnitude-based Token Selection (GMTS) method to quantify token importance, which leverages the entropy-gradient connection to approximate gradient-magnitude rankings for token selection. We find that training on the top 20% tokens ranked by GMTS consistently outperforms entropy-based token selection across three reasoning domains and various model sizes, suggesting that GMTS provides a more fine-grained estimate of token contribution for RLVR training.

Large language models (LLMs) finetuned for specialized domains represent crucial high-impact applications. Inference-time alignment improves safety degraded from specialization finetuning without requiring substantial computational resources, complementing finetuning-based methods with an easy-to-use, plug-and-play solution. However, existing inference-time methods fail to reliably improve safety without disrupting domain capability. We identify the root cause as complementary expertise orthogonality: specialized base models and general-domain guidance models have orthogonal competencies, making the guidance signal unreliable for specialized generation. This primarily manifests as stop token interference, where the guidance model's tendency toward continuation overrides the base model's decision to stop, burying correct answers under guidance-induced continuation. To address this problem, we propose CREST, an inference-time alignment method that steers base model hidden representations using safety directions extracted from a guidance model of any family, avoiding token-level structural limitations entirely. CREST improves safety where specialization has weakened it while preserving both domain-specific capability and the safety of already well-aligned models, outperforming baselines by up to 22.2% on safety benchmarks. Our code is available at: https://github.com/DecayingSeart/CREST.

Brazilian Portuguese remains under-served by open language models, and the few that exist are difficult to reproduce and are often compared without measures of uncertainty. We release Manacá-1B, an open decoder-only model of 1.72 billion parameters trained from scratch for Brazilian Portuguese with a fully containerized, reproducible pipeline. The pretraining is stable, with zero skipped or NaN steps and self-recovering loss spikes, and we release its full log and dynamics. We evaluate the model against nine open baselines on four Portuguese benchmarks under a single harness. Every comparison reports a standard error and a paired significance test, and the harness is validated against previously published numbers. On last-word prediction Manacá-1B is the strongest model below the 7B scale, exceeding both Tucano-1b1 and Tucano-2b4 on LAMBADA-PT with large paired margins; it is competitive on commonsense completion and near chance on multiple-choice reasoning, as are all small base models. Along the way we document a concrete evaluation pitfall: converting a SentencePiece tokenizer with case-folding normalization to the HuggingFace fast format silently drops the normalizer, routing every capitalized token to byte-fallback and depressing scores in a way that is invisible in aggregate metrics. The uncorrected tokenizer lowered LAMBADA-PT accuracy from 45.3 to 25.0; we quantify the effect and provide a one-line fix that reproduces the training tokenizer exactly. Code, raw training and evaluation logs, per-example prediction vectors, the model weights, and the corrected tokenizer are released so that every number in this paper can be recomputed.

Introduction: NICE guidelines provide evidence-based recommendations for clinical care but remain largely in unstructured natural language. Existing approaches to converting them into computable representations often focus on individual diseases, require substantial manual encoding, and do not scale. Large language models (LLMs) may enable much of this translation to be automated. Methods: We present an end-to-end approach that converts textual clinical guidelines into executable models capable of generating explainable, patient-specific recommendations. A stepwise LLM-based transformation with in-context examples produces human-inspectable intermediate artifacts. We apply the approach to NICE pancreatic and lung cancer guidelines, use expert review to assess rule alignment, and evaluate the executable pancreatic cancer model on 20 patient vignettes. Results: Expert review showed strong alignment between the source guidelines and generated executable models. Most discrepancies were partial omissions rather than incorrect logic, while hallucinated or fundamentally incorrect rules were rare. On the patient vignettes, the executable model achieved an F1 score of 82.5%. Conclusion: LLMs can transform natural-language NICE guidelines into interpretable, executable models that preserve guideline structure, support transparent inspection and modification, and generate patient-specific recommendations. These findings demonstrate the feasibility of scalable automated generation of computable clinical guidelines.

We ask whether large language models (LLMs) can design effective algorithms for well-specified operations research (OR) problems. We study inventory control, queueing network control, and assortment optimization. We evaluate two levels of LLM use: at level 1, the model receives one problem instance and returns a solution for that instance; at level 2, it receives only the problem class description and broad parameter ranges, and returns an algorithm that maps instance parameters to solutions. Human input is minimal: we give one untuned prompt that describes the problem, and the model has access to a Python sandbox tool with a fixed compute budget. The strongest model we test, gpt-5.6-sol, matches or outperforms the best existing method on almost all evaluated instances. This holds even at level 2, where the returned algorithm is fixed before seeing the evaluation instances. Performance also improves sharply across models released less than eight months apart, suggesting that this capability is moving quickly. Thus, for the well-specified operations problems we study, a single untuned LLM query can already produce algorithms competitive with specialized methods. These results suggest that frontier LLMs can be a serious empirical baseline for algorithm design in well-specified OR problems.

User behavior simulation with large language models~(LLMs) is increasingly used to support multi-agent ecosystem simulation. Existing simulators typically rely on static user profiles inferred from historical observations, which become inadequate in socially intensive environments such as live streaming where interaction dynamics continuously reshape user behavior. We propose \textbf{LiveSim}, an LLM-based framework for live-stream ecosystem simulation. It represents users as editable behavioral hypotheses and progressively refines them through trajectory-grounded interactions, where discrepancies between simulated and observed trajectories reveal missing environmental shaping effects. These signals are further extracted as transferable environment-behavior patterns and accumulated in a collective behavioral memory to improve user-level behavioral fidelity and support ecosystem-level simulation. Experiments on real-world live-stream risk-control data validate the effectiveness of LiveSim in improving user-level behavioral fidelity and enabling ecosystem-level analysis of risk evolution and platform intervention effects.

Current large language model development relies on massive, often non-permissible datasets, creating a high barrier for researchers committed to open-source and ethically sourced data. We introduce Mimir v1, a 1-billion-parameter language model based on the Hierarchical Reasoning Model (HRM) architecture, that is trained from scratch and delivers highly competitive performance for English and sets a new state of the art for Danish using only permissible post-training data. Trained on a mixture of 161 datasets, Mimir v1 outperforms the original HRM-Text 1B and competes with larger frontier models like Qwen 3.5 4B and Gemma 4 E2B, tested across 20 benchmarks for English, Math & Code and Danish. The model is available on the Hugging Face Hub: https://huggingface.co/danish-foundation-models/DFM-Mimir

We present Multi-Agent Reasoning and Coordination (MARC), an open-source framework that replaces monolithic LLM prompting with deterministic multi-agent orchestration for clinical reasoning. MARC coordinates role-specialized agents for extraction, reasoning, answer generation, and evaluation, with explicit context passing and traceable intermediate outputs, enabling stage-wise failure attribution. We additionally introduce a Decomposer module that generates task-specific agent prompts from a plain-language description, eliminating manual prompt engineering. The framework supports both API-based and local CPU-compatible deployments and is entirely configurable via YAML, without code modifications. MARC is designed to be model-agnostic, interpretable, and accessible to clinical domain experts without programming expertise. The full framework is available at https://github.com/Penn-RAIL/MARC-v1.

We ask whether language-model pre-training can be decomposed into smaller, independently trainable jobs that can later be recomposed into a coherent larger model. We introduce Mixture of Training (MoT), a scaffolded modular pre-training procedure that partitions a target Transformer into contiguous layer blocks, trains each block inside a frozen pretrained aligner scaffold, and then recomposes the trained blocks with an optional short end-to-end adaptation pass. On a 1.3B-parameter Gemma-style model trained on C4, MoT provides a small-scale proof of mechanism: independently trained depth slices can be recomposed into a usable language model, and a quality-parity schedule reaches the same reported perplexity as the monolithic baseline. This parity setting processes more aggregate tokens and has a shorter idealized layer-equivalent critical path after aligner preparation; its effective compute advantage depends on reusing the aligner across runs. We therefore present MoT not as a general replacement for monolithic pre-training, but as a small-scale framework for studying whether scaffolded sub-runs can act as reusable training units.

The extent to which large language models for code rely on memorization over genuine understanding remains highly debated. While current literature frequently reports widespread memorization, evaluating the underlying probing techniques across dense architectures reveals a severe breakdown in their utility at scale. Traditional encoder-style probes using perturbations such as synonym fuzzing or dead-code insertion struggle to expose memorization in scaled models, even on known-contaminated benchmarks, and decoder-style probes that rely on log probabilities show similar performance degradation. The specific mode of failure for these probes, particularly why such techniques disrupt smaller models but fail to impact larger ones, motivates us to untangle representation load from memorization rather than treating them as a single phenomenon. By applying invertible mathematical transforms to numeric problems, we isolate these two factors and reveal that scaled encoders successfully absorb substantial representation load while still converging on the correct family of solutions. In practical software engineering, this ability to adapt to varying surface forms is what truly matters for usability and generalizability in LLM and agentic applications. Whether a specific solution was seen during training becomes a much less pressing question because although memorization inflates scores on contaminated benchmarks, factoring out representation load makes it debatable how much we should truly care if a functional answer was originally memorized. Future evaluations must therefore be built around separating these phenomena rather than relying on methodologies that quietly entangle them.

Medical AI has demonstrated specialist-level diagnostic accuracy, yet these capabilities remain largely inaccessible in resource-constrained rural settings where bandwidth is scarce, compute is limited, and clinical decision-making requires integrating heterogeneous modalities. We introduce a cloud--edge collaborative architecture that addresses these constraints: lightweight, domain-specific models on the edge transform raw medical data into compact structured outputs, while a cloud LLM synthesizes these outputs into clinical summaries. An LLM-based orchestrator dynamically selects diagnostic tools based on patient context, promoting comprehensive modality coverage without processing irrelevant inputs. We evaluate on 20 multimodal clinical cases spanning cardiac, obstetric, trauma, and screening scenarios under three simulated network profiles (500,kbps--5,Mbps). The hybrid system achieves 98--99% diagnostic tool recall with 92--96% precision, matches or exceeds cloud-only baselines on clinical accuracy, and maintains bandwidth-invariant latency (25--35,s) at 4--15x lower token cost. These results highlight the role of architectural design in enabling efficient multimodal integration and improving factual grounding compared to cloud-only approaches under deployment constraints.

Retrieval-Augmented Generation (RAG) is widely used to improve the performance of Large Language Models (LLMs) in answering user queries. Existing privacy research on RAG has focused on preventing unauthorized users from accessing sensitive data. However, another important problem that is often overlooked in RAG privacy research is that external generators have access to the query and the retrieved documents, which may contain confidential information that could potentially be misused or accessed for unintended purposes. In this paper, we introduce the Sensitive Entity Alias Generator (SEAG), a privacy-preserving framework that empowers users to utilize powerful third-party generators without disclosing sensitive information. SEAG introduces a lightweight model that locates sensitive entities, generates corresponding aliases, and constructs an entity replacement table. The table is used to replace sensitive words in the user's query and in the retrieved documents before they are forwarded to an external generator. For this purpose, two datasets were constructed: one for fine-tuning SEAG models to generate entity replacement tables, and another for evaluating the entire SEAG framework. The experimental results demonstrate the success of the SEAG framework. As for the User metric, which measures the ability of the model to provide a correct response to the user while hiding sensitive information from the external generator, all SEAG models achieved over 80% accuracy. Additional analysis further evaluated the ability of SEAG models Qwen-3, LLaMA-3.2, and Phi-4 to hide all sensitive entities within given documents. The results show good performance with total accuracies of 77.83%, 76.73%, and 74.91%, respectively.

Large language models can extract richer signals from financial news than fixed sentiment lexicons, and recent work has explored feeding such signals into portfolio construction. We study an uncertainty-aware construction that feeds model-predicted risk -- decomposed into aleatoric and epistemic components -- directly into the covariance matrix of portfolio allocators, rather than treating portfolio risk as fixed or adjusting only expected returns. We evaluate the pipeline on Russell 2000 equities under three stock-selection regimes: a pure-alpha trigger that isolates abnormal stock moves not explained by macro indicators, a pure-beta trigger that captures macro-indicator moves before the stock itself fires, and a beta trigger in which both channels agree. Across the full holding-period grid, the separated pure-alpha and pure-beta legs usually dominate the beta intersection on Sharpe and return. Two horizons are especially informative. At one day, pure beta can work under low and moderate transaction costs because it captures immediate lead-lag spillovers from liquid macro and sector indicators into exposed small-cap stocks, but this advantage disappears at 100 bps when turnover and microstructure noise dominate. At 40 days, pure beta works for a different reason: slower macro repricing overtakes the firm-specific pure-alpha channel. The strongest conservative row is pure beta with GPT-4o mini sentiment, a Student-t target, a 40-day holding period, and risk parity allocation, reaching Sharpe 2.33 at 100 bps. The results suggest that stock-selection regime and allocator choice matter at least as much as the sentiment model, and that separating firm-specific and macro-exposure triggers is more informative than requiring both to fire simultaneously.

Large language models are increasingly trained and deployed with long contexts that span documents, code repositories, and interaction histories. This scaling reflects the implicit assumption that training on longer contexts will only help the model by exposing it to richer evidence. We challenge this view by studying how the context window shapes a model's mode of learning, shifting it between parametric internalization and contextualization. We propose the Information Abundance Paradox, which hypothesizes that abundant relevant information in the training context can reduce the incentive to encode that information parametrically, thereby increasing reliance on context. In pretraining with long documents, increasing the context window improves language modeling, natural language understanding, and closed-book MCQA only up to an intermediate optimum, after which performance consistently declines. In supervised fine-tuning, more task-relevant train-time context improves performance with supporting context, but reduces robustness when context is absent or misleading at test time. Our analysis suggests that this behavior arises when longer context provides a lower complexity solution. Mechanistically, training with informative context shifts gradient pressure from feed-forward networks, often linked to parametric knowledge, toward attention modules, and causal interventions show that this shift increases reliance on context during inference. Overall, these findings support the Information Abundance Paradox and suggest that scaling toward near-infinite context is not simply a matter of supplying more data, even when high-quality long-context data is abundant.

We present the first systematic study of Massive activations (MAs) in layer-interleaved HLA LLMs and uncover two architecture-aligned morphologies: MAs consistently spike immediately before full attention layers, forming pre-attention spikes (PAS), and can persist through intervening linear attention layers, giving rise to inter-spike plateaus (ISP). As full attention becomes denser, successive PAS become increasingly connected through ISP, ultimately recovering the stable MA morphology of full attention LLMs. We establish the recurrence of this organization across five linear attention architectures, six hybridization configurations, five data domains, and representative open-source hybrid models spanning 1.2B to 397B total parameters. Controlled pretraining of GDN-based hybrids at scales up to 1.3B shows that both morphologies emerge early and respond asymmetrically to output gating: full attention output gating strongly attenuates their absolute magnitudes without eliminating their layerwise organization, whereas removing GDN gates yields comparatively modest amplification. Mechanistically, our systematic-outlier analysis supports a shared lifecycle account governed by the timing of MA cancellation. PAS follows a localized write-sink-cancel process, while the extended persistence of ISP is consistent with delayed cancellation. At the full attention limit, this account recovers the stable MA morphology characteristic of full attention LLMs. Our code is available at https://github.com/StartluxLabs/Massive-Activations-HLA.

Small Language Models (SLMs) have emerged as a more efficient alternative to traditional Large Language Models (LLMs), offering promising potential in resource-constrained scenarios. Existing approaches to building SLMs typically follow two paths: training compact models from scratch, or compressing larger pre-trained models using methods such as pruning, quantization, or distillation. As language models become increasingly integrated into real-world applications, ensuring their trustworthiness has become a critical concern. However, how to build trustworthy SLMs remains an underexplored question. In this work, we present a comprehensive evaluation of SLM trustworthiness across multiple dimensions, including fairness, robustness, privacy, and ethics. We first examine the effects of pruning and quantization, and find that quantization is significantly more effective in preserving trustworthiness compared to pruning. More importantly, we demonstrate that compressing a reliable large model via quantization can produce SLMs with superior trustworthiness and adaptability compared to using small models trained from scratch. Furthermore, knowledge distillation from trustworthy teacher models can further enhance the reliability of SLMs. We hope our findings provide practical guidance and a foundation for future research into the development and deployment of trustworthy small language models.

State-aligned distortion has been documented in China-origin text-based large language models (LLMs), but whether, and in what form, it arises in multimodal systems has not been systematically examined. We construct a balanced benchmark of 200 core entries spanning ten politically sensitive topics, plus a seven-variant visual-abstraction probe, and run nine vision-language models (VLMs), seven China-origin and two non-China, across four elicitation paradigms and two prompt languages, yielding 21,708 trials. Each response is audited on six dimensions -- explicit refusal, information integrity, visual grounding, state-aligned framing, language consistency, and response length -- by two independent frontier LLM judges, validated against three human experts on a 200-trial sample. Measuring each dimension separately lets us decompose multimodal censorship into individual signals rather than a single refusal-based score; in particular, refusal and framing are measured independently, so a model can stop refusing while still reframing. We find that (i) Chinese-language prompting roughly triples the odds of state-aligned framing, within every model; (ii) China-origin models reframe more than non-China models (direction robust across judges and human raters; magnitude 1.6--3.2x); (iii) the effect is strongest in text-only political commentary (36.5%) and is gated by recognition of the depicted subject rather than pixel detail, persisting even at silhouette for iconic images; and (iv) across four Qwen generations, state-aligned framing rises while explicit refusal falls: censorship migrates from a visible act (refusal) to an invisible one (fluent reframing). We argue this shift to invisible reframing is fundamentally a problem of human-AI interaction: it removes the very signal users rely on to recognize that information has been withheld.

Generating actionable financial advice from business records demands that models integrate numerical reasoning, domain knowledge, and sound judgment, while avoiding recommendations that could harm the business. Direct supervision is difficult: historical decisions are not necessarily optimal, and high-quality free-form labels are expensive to obtain. We formulate financial advice generation as a reinforcement learning problem and fine-tune an open-weight language model using Group Relative Policy Optimization (GRPO). Our reward is an LLM-as-a-judge rubric that scores each recommendation across multiple binary dimensions of advice quality, augmented with a safety gate for harm prevention. Since LLM-based evaluation alone cannot confirm whether improvements reflect genuine business value rather than adaptation to the judge, we complement it with a judge-independent audit based on a standard doubly-robust Conditional Average Treatment Effect (CATE) estimator. Under this observational off-policy audit, our trained LLM achieves approximately twice the estimated gross-profit lift of the strongest evaluated commercial baseline ($0.0228$ vs.\ $0.0104$), together with the lowest downside rate and the least negative tail risk of any policy evaluated. Notably, the two evaluations do not rank the baselines identically: the untrained base model places last on the judge rubric but second on the causal audit, indicating that the audit captures a signal the judge does not. Our results demonstrate that GRPO with a finance-grounded reward signal can produce substantially more useful business recommendations than commercial LLMs, and that a judge-independent causal audit is a valuable complement to, rather than a confirmation of, LLM-as-a-judge assessment in financial NLP.

We study whether the localized numerical operations and integrative judgments of financial analysis benefit from the same form of LLM specialization. Larix maps a 16-lens European listed-real-estate analysis framework to eight lens-aligned specialists; we compare a frontier LLM under monolithic versus specialist-decomposed prompting while holding the model, source evidence, task instructions, output schema, and scoring fixed. Across 19 firms spanning seven regulatory wrappers, decomposition improves the numerical-task aggregate by 15.8 percentage points but does not reliably improve, and can reduce, performance on judgment tasks, a pattern stable across four frozen-template dispatches; a single-agent control given the complete framework does not reproduce the numerical gain. Post-training Qwen3.5-9B with GRPO using task-aligned structured rewards then raises the development-split score by 12.0 points and the judgment aggregate by 14.2 points, with gains on all four sub-ceiling tasks; the gains transfer to unseen firms (+15.2 points overall; +40.4 on covenant stress) and to unseen regulatory wrappers (+4.3), with positive transfer on all three anti-memorization splits. Prompt-level decomposition thus improves modular numerical execution, whereas targeted parameter adaptation improves integrative financial judgment.

Tokenizer vocabulary size is a foundational design choice in large language model (LLM) infrastructure, yet it is typically fixed at training time based on convention rather than deployment analysis. We show that the cost-optimal vocabulary is not a constant but a function of the serving regime. We formalize total deployment cost as $C_{lifecycle}(V) = C_{train}(V) + λ\cdot C_{infer}(V, B)$, where $λ$ is inference volume and $B$ is the serving batch size. Through controlled experiments on two GPU families spanning the memory-bound to compute-bound regimes (A10G, ridge $\approx$ 117 FLOP/byte; A100, ridge $\approx$ 183 FLOP/byte), we demonstrate: (1) the inference-optimal vocabulary shifts 16x with serving batch, from 32k at $B=1$ to 524k at $B=64+$, driven by amortization of the $V \times d$ unembedding matrix read; (2) at 1.3-2.3B model scale, quality (bits per byte, BPB) is optimized at $V=65$k, confirming scale-dependent vocabulary preference; (3) the lifecycle-optimal vocabulary diverges from training-optimal by up to 16x for production deployments. Quality is approximately invariant across the optimal range ($<$2% BPB spread), making vocabulary a pure systems optimization with no quality penalty in the measured range. Our results provide actionable capacity planning guidance: on-device deployments ($B=1$) should use $V \approx 32$k; datacenter serving ($B \geq 64$, $λ\geq 10$) should use $V \approx 131$-262k.

Edge LLMs must preserve inference continuity when a user hands over between edge nodes, requiring key-value (KV) cache transfer to the target node. However, simultaneous handovers saturate the backhaul, preventing full cache delivery within the mobility-imposed transfer window. Rather than allocating bandwidth as if all cache entries were equally valuable, we order each user's KV cache by importance and transmit only its most informative fraction, turning token-level sparsity into communication savings. We cast the transfer as a multi-user backhaul allocation problem that maximizes average accuracy across users. Each user's partial-cache accuracy serves as its utility: a sigmoid that fits measurements on the RULER benchmark with $R^2>0.99$ across models and context lengths. Because importance ordering front-loads the high-value entries, the concave region of the accuracy curve spans nearly the entire cache. Our proposed allocator keeps served users within this region, making each per-slot allocation problem convex. The optimum is derived via a closed-form weighted water-filling solution that generalizes information-theoretic water-filling and enables online scheduling. The proposed allocator attains over 93.7% average accuracy in a 500ms transfer window, within 0.5pp of the full-cache ceiling, and reaches 98.2-99.5% of a clairvoyant upper bound.

Large language models are increasingly being deployed in governmental settings, yet few existing evaluation frameworks jointly reflect the values of public administration and the linguistic requirements of non-English contexts. We present the "Grip on LLMs" framework, a systematic evaluation suite for Dutch governmental use developed in collaboration with domain experts from a major Dutch municipal organisation. Through an advisory board process, user research, and a survey of the users of a civil-servant chatbot, we identify six evaluation dimensions (factuality, honesty, social bias, energy consumption, cost, and training data transparency) and operationalise them into a benchmark suite covering more than 30 multilingual and Dutch-specific models. Our results reveal that no single model excels across all dimensions, and that trade-offs are unavoidable: higher quality consistently comes at greater environmental impact and financial cost, while bias remains largely independent of both. We further find that factuality (whether a model answers correctly) and honesty (whether a model acknowledges what it does not know) are governed by distinct properties, with high factuality not implying high honesty. To make these findings actionable for non-technical audiences, we release a publicly accessible, user-friendly model overview designed for the full range of stakeholders involved in governmental LLM selection, from engineers to policymakers.

Large multi-call LLM systems pose a scientific problem that query-level routing does not capture: the value of a model depends on where it enters a dependent computation and on the deployment that surrounds that call. Existing routers typically decide \emph{where} to spend a stronger model while treating the benefit of the substitution itself as known. We separate these two decisions through a predicate-action factorization and evaluate it in controlled solve-merge-verify workflows spanning 8-64 solve calls and four three-tier model ladders. The resulting evidence reveals a consistent principle beneath apparently conflicting outcomes. On numeric frequency counting, all-strong reduces RMSE from 4.818 to 1.538 in the Mixed Qwen/GPT ladder, whereas the average Qwen-only ordering reverses. Input-matched interventions further show that the same medium-to-strong action has sharply different value across roles and scales. A semantic task-and-contract shift reverses the Mixed ordering again, while allocation ablations distinguish useful sparse placement from under-coverage and indiscriminate escalation. Together, these results establish model substitution as a deployment-conditioned action rather than a property implied by a tier label, and they provide a practical sequence for large-scale workflow routing: calibrate the action, resolve its role-conditioned effect, and then optimize its placement.

The carbon footprint of any deployed Large Language Model (LLM) accumulates during inference, where repeated use of the model substantially exceeds the one-time cost of fine-tuning. Yet most efficiency interventions target either pre-training scale or post-hoc compression. We ask whether folding a calibrated, differentiable energy surrogate into the fine-tuning objective can produce inference behavior that gains task accuracy at zero or near-zero carbon cost, a break-even configuration. We propose a joint loss mechanism with a per-model carbon-emission parameter, a linear surrogate over parameter norm, FLOP proxy, and a memory proxy, fit from on-hardware energy profiling. We fine-tune three architecturally distinct families: Gemma-2 2B, Llama-3.1 8B, and Qwen-2.5 14B, and evaluate inference F1 and CO$_2$ emissions on three MMLU subjects: abstract algebra, philosophy, and formal logic. We discover from several outcomes that the carbon term behaves as either harmful interference or beneficial regularization depending on the task structure. We position calibrated carbon-aware fine-tuning as a lightweight, drop-in regularizer with a non-empty but model and task-dependent break-even region. This is an ongoing work, and we will release our codebase soon.

Large Language Models (LLMs) have significantly revolutionized the modern society with numerous advanced interactions between humans and AI agents, whereas the usage of most large language models including ChatGPT are not friendly open-sourced and must require the users paying a lot for such AI services continuously. Therefore, deploying open-sourced large language models on local servers can be considered as an efficient approach to design and implement creative embodied AI algorithms with lower cost and more stable free usage. Inspired by this ordinary motivation, we originally propose and implement the "Socratic Models-ChatGLM", which is a well-performed algorithm for multi-modal interactive control of robotic arm based on offline large language models via the facile PyBullet platform, even presents extraordinary potential to address complicated text-image integrated multi-step long-horizon robotic manipulation tasks.

No single large language model (LLM) is optimal across all queries and budget constraints, making model routing essential for cost-effective deployment. Existing routers adopt diverse formulations and implementations, making fair comparison and extension difficult. We present a unified formulation of LLM routing as a sequential decision process characterized by five components: context encoders, model encoders, scoring functions, decision rules, and learning signals, covering single-turn, multi-turn, and personalized routing. Based on this formulation, we develop an automated pipeline for constructing routing supervision and evaluating routers jointly on response quality and inference cost. The resulting benchmark, xRouteBench, spans generic LLM, memory-augmented, vision, time-series, and personalized routing tasks. We further introduce LLMRouter, an open-source modular infrastructure with more than 16 representative routers. Our empirical study shows that learned routers outperform the strongest fixed-model baseline by 14.6% relatively, lightweight routers become more competitive under tight cost constraints, and user-conditioned routing consistently improves personalization.

Large language models are increasingly used to read markets, assess risk, and allocate capital. However, reported results for LLM trading agents can be inflated by look-ahead leakage, optimistic execution, and risk mandates that are described but not enforced. We present OpenPM, an auditable point-in-time evaluation framework for LLM portfolio-management agents. In OpenPM, an agent manages a $1M long-only book over the S&P 500 universe using market data at five-minute intervals. Every record visible to the agent must be available at the decision time. Natural-language risk mandates are converted into typed constraints and enforced on the executed portfolio. Each run produces audit artifacts, including a contamination certificate, a cost-sensitivity curve, and a constraint-adherence report. We also build a reference agent named the tiered allocator, where typed analysts score candidates, a constructor LLM proposes weights, and a deterministic critic guarantees feasibility. We isolate constructor behavior by capturing analyst evidence once and replaying it across constructor models. In our short-window case study, stronger constructors show modest and model-dependent gains over equal weighting on the same pool, but analyst quality matters more than constructor choice, and turnover is the main cost driver. All returns are upper bounds on a single frozen window without market impact, not validated alpha.

Large language models (LLMs) can leave small stylistic traces in text written with their help. The most discussed is the em-dash (U+2014), especially the unspaced form word---word, which is normal in typeset English prose but unusual in U.S. press writing, where AP style calls for spaced dashes. This study asks whether that trace is measurable in congressional press releases. In a preregistered design (OSF: 10.17605/OSF.IO/U5NEY), 146,239 scraper-sourced releases from 480 House and Senate offices (2021-2025, the open congress-press dataset) were analyzed: density of unspaced prose-form em-dashes per 1,000 characters of cleaned text, Poisson/negative-binomial models with a length offset, clustering by office. Density stayed within 0.10-0.12 per 1,000 characters through 2021-2024, then rose to 0.217 in 2025, more than twice the four-year baseline; the share of releases with such an em-dash rose from ~13% to 24.8%. The primary frequency ratio (2023-2025 vs 2021-2022) was 1.55 (95% CI 1.28-1.93; exact registered cut-off: 1.528), just above the prespecified 1.5x threshold. The rise was net-new (hyphen density stable), held within authors (75.6% of 262 continuous offices increased; p ~ 1e-16) and in a closed panel of 224 offices, and survived falsification tests: three placebo cut-offs were null, the pipeline showed no step at the 2024/2025 boundary, and continuing offices carried the rise. A segmented regression finds no step at the ChatGPT cut-off but a clear post-period acceleration; the 2025 rise is symmetric across parties and chambers. Because the registered validation gate was formally breached, the full preregistered decision rule was not met; the interpretation (broad diffusion of LLM-assisted writing as the models matured) is offered as exploratory. The em-dash remains a population-level marker, not a per-release authorship detector, and the design supports no causal claim.

Standard Large Language Models (LLMs) execute layers sequentially. Dynamic layer routing, i.e. search for a different execution path through layers involving layer repetitions, skips and other moves, can improve performance. Existing routing approaches often require updating model weights, running expensive search loops per test instance, or demand ground-truth labels during inference. In this work, we propose Markov Chain Routing of Transformer Layers (MACRO), a framework that learns task-specific routes over LLM architectures without modifying underlying parameters. MACRO models layer routing as a context-dependent Markov policy conditioned on layer indices, computation budget phases, directional displacements, and operator context, supporting skip, repeat, and residual hidden-state addition operations. The Markov route distribution is updated via feedback on training data and decoded using a top-k Viterbi algorithm to isolate high-probability candidate programs. We evaluate MACRO across diverse reasoning and knowledge benchmarks on multiple open-weight LLMs. MACRO achieves a +5.0% average accuracy improvement over the unrouted baselines, with largest gains on small models. We outperform the best dynamic routing approach Dr. LLM by +7.2%, while reducing route-search time 9.4x (from 14.8 to 1.6 hours). Our code is publicly available at https://github.com/Batorskq/MACRO.

This paper presents a conceptual framework for developing an electronic explanatory dictionary of the Tajik language using large language models (LLMs). The relevance of the work stems from the absence of a comprehensive digital lexicographic resource for Tajik that is comparable in functionality to dictionaries for high-resource languages, and from the limited adaptation of modern natural language processing technologies to low-resource language systems. Based on a systematic survey of existing linguistic, statistical, and corpus resources, we propose a dictionary architecture that integrates modules for morphological analysis, lemmatization, semantic clustering, and dictionary entry generation using LLMs. The choice of subword tokenization is justified by the agglutinative nature of Tajik morphology and its high morphological variability, along with a parameter-efficient fine-tuning (PEFT) strategy suitable for limited annotated data. The novelty of the work lies in proposing the first holistic conceptual architecture of an explanatory dictionary for Tajik that unifies classical lexicographic methods, language statistics, and generative capabilities of LLMs into a single system. The practical significance of the study is the formation of a methodological foundation for developing a full-featured electronic dictionary that can serve both as a lexicographic tool and as a core resource for machine translation, automatic summarization, sentiment analysis, and other applied NLP tasks. The paper is intended for specialists in computational linguistics, lexicography, and developers of natural language processing systems working with low-resource languages.