Computation and Language · Cross-listed from cs.LG
The Geometry of Learning: Read-Write Asymmetry Across Training
Valeria Ruscio and Keiran Thompson
Pretraining and
alignment update the same
transformer weights, and we ask in which directions each stage changes them. The residual stream gives every
matrix a shared coordinate system, in which a weight change can be compared with the principal subspace of the activations the
matrix reads and with the prediction subspace of the unembedding. In this frame,
alignment updates to the query and key
matrices are enriched in the principal input subspace, while its updates to the output
matrices are close to random relative to the prediction subspace, and the same signature appears in late pretraining and under every objective we apply to one checkpoint, so it is a property of the training stage rather than of the objective. The gradient of a linear map is a sum of outer products $\delta_t a_t^\top$, and the residual stream enters this product through the activation $a_t$ for
matrices that read from it and through the upstream gradient $\delta_t$ for
matrices that write into it. Read gradients therefore inherit the spectrum of the activations, which is the same under every objective, while write gradients inherit the geometry of the upstream gradients, which point into the prediction subspace only early in pretraining, when the output
matrices acquire their prediction geometry. The raw read gradient follows the activation spectrum, the optimiser attenuates the concentration it inherits, and cancellation between
tokens sets the scale of the update but not its direction. A write direction derived from the unembedding still changes predictions after
alignment, and adaptation restricted to any single group of
matrices remains effective, so the signature describes where training places its updates, not which
matrices can carry them.
Read the full abstract ▾
Computation and Language
Scaling Accessible Mathematics on arXiv: HTML Conversion and MathML 4
Deyan Ginev, Brian Caruso, Bruce Miller et al. (+2 more)
We report on the ongoing development of arXiv's HTML Papers offering, available on every new TeX/LaTeX submission since its initial release in 2023. The main highlights from 2025 and early 2026 are: (i) community-driven improvements to HTML fidelity and service health, with roughly half of 6,000 user reports resolved; (ii) corpus-scale conversion work aimed at 90% error-free HTML (currently 75%); (iii) initial MathML 4 Intent annotations for accessible
speech output; (iv) an in-progress Rust port of LaTeXML, reducing compute costs and enabling faster previews on submission. The arXiv HTML Papers project remains experimental, but is gradually maturing as we better understand the needs of arXiv's readers and the technical opportunities presented by new standards and by advances in programming
languages and AI.
Read the full abstract ▾
All authors Deyan Ginev, Brian Caruso, Bruce Miller, Jeff Sank, Jacob Weiskoff
Computation and Language
PQR: A Framework to Generate Diverse and Realistic User Queries that Elicit QA Agent Failures
Yunan Lu, Luigi Liu, Omar Yahia et al. (+2 more)
Evaluating LLM-based
agents remains challenging because identifying meaningful failure cases often requires substantial human effort to design realistic test scenarios. Prior works primarily focus on automatically discovering
agent failures induced by
adversarial users, while overlooking queries with real user intents that also trigger
agent failures. We introduce PQR, a framework that not only surfaces
agent failures with respect to specific objectives (e.g., helpfulness, safety, etc.) but also resembles real users' intents. PQR operates through an iterative interaction between two complementary modules. The query refinement module performs rewrites to explore diverse query variations, while the
prompt refinement module uses prior feedback to derive new objective-violating strategies and realism policies for refining
prompts, which in turn generate failure-triggering yet realistic queries. We evaluate PQR on detecting an e-commerce QA
agent's unhelpful responses. Our method uncovers 23% - 78% more unhelpful responses, and our generated queries are more diverse and realistic compared to previous methods.
Read the full abstract ▾
All authors Yunan Lu, Luigi Liu, Omar Yahia, Arpit Sharma, Zhou Yu
Computation and Language · Cross-listed from cs.LG
Symphony for Speech-to-Text: Supporting Real-Time Medical Voice Interfaces
Arne Nix, Robert James, Lasse Borgholt et al. (+6 more)
After decades of use in dictation and, more recently, ambient documentation,
speech is emerging as a primary modality for interacting with technology and AI in healthcare. Yet medical
speech recognition remains difficult: systems must capture specialized terminology, resolve contextual ambiguity, and render measurements, abbreviations, and clinical shorthand precisely. Existing solutions are typically optimized either for general-purpose transcription or narrow dictation workflows, limiting their reliability in safety-critical settings and their usefulness for broader clinical workflows. We introduce Symphony for
Speech-to-Text, a medical-grade
speech recognition system for real-time streaming and batch file-based clinical use. Symphony decomposes the transcription process into specialized components for recognition, formatting, and contextual correction to optimize medical term recall while producing clinically structured text in real time and adapting across use cases. Evaluations on public
benchmark and medical
speech datasets show that Symphony substantially outperforms state-of-the-art systems in clinical settings while matching or exceeding them in general-domain settings, suggesting robust generalization rather than overfitting. We release a clinical
benchmark dataset to support reliable validation and further progress in medical
speech recognition. Symphony is available through a production-grade API for live dictation, conversational transcription, and batch audio file processing.
Read the full abstract ▾
All authors Arne Nix, Robert James, Lasse Borgholt, Anna B. Ekner, Lana Krumm, Julius Severin, Dan Engel, Lars Maaløe, Jakob Havtorn
Computation and Language · Cross-listed from cs.HC
LLMs in Qualitative Research: Opportunities, Limitations, and Practical Considerations
Henry Salgado, Meagan R. Kendall, Martine Ceberio et al. (+1 more)
This paper examines the opportunities, limitations, and practical considerations associated with the use of large
language models (LLMs) in qualitative research. Drawing on a multidisciplinary perspective that combines expertise in qualitative methods and explainable AI, the paper argues that responsible integration of LLMs into qualitative workflows requires researchers to engage critically with a curated set of technical parameters, that is, context window constraints, temperature and top-p sampling settings, user and system
prompt design, and model documentation in the form of system cards. The paper situates these considerations within the epistemological commitments of qualitative research, including reflexivity, positionality, and interpretive judgment, and discusses how the opacity of contemporary LLMs differs from earlier natural
language processing tools such as topic models and lexicon-based sentiment analyzers.
Read the full abstract ▾
All authors Henry Salgado, Meagan R. Kendall, Martine Ceberio, Alexandra Coso Strong
Computation and Language · Cross-listed from cs.HC
Alignment Drift in Long-Term Human-LLM Interaction: A Mechanism-Oriented Framework
Xintong Yao
Long-term interaction with LLM-based systems may produce
alignment drift: a gradual process in which system outputs become less constrained by the user's current message and more shaped by prior interaction history, while still appearing helpful, coherent, and responsive. This process is difficult to detect because the user's subjective experience may improve as the system becomes more familiar, useful, and attuned. Existing research on human-LLM interaction has largely focused on short-term task performance, isolated outputs, or single-instance
alignment problems, leaving slow and cumulative interaction-level dynamics undercharacterized. This paper proposes a mechanism-oriented framework for describing
alignment drift. The framework defines the distinction between signal A and signal B, explains how drift develops through feedback loops and sub-pattern selection, divides the process into three interactional regimes, and identifies boundary conditions for controlling drift. By framing
alignment drift as a recursive interactional process rather than an isolated model-side failure, the paper provides a conceptual basis for studying long-term human-system interaction.
Read the full abstract ▾
Computation and Language
The Scaling Laws of Skills in LLM Agent Systems
Charles Chen, Qiming Yu, Yuhang Gu et al. (+12 more)
As
agent systems scale, skills accumulate into large reusable libraries, yet their scaling laws remain poorly understood. Across 15 frontier LLMs, 1,141 real-world skills, and over 3M routing or execution decisions, we identify two coupled laws. Routing law: single-step routing accuracy decays logarithmically with library size ($R^2{>}0.97$ for all models), with errors progressing from local skill competition to cross-family drift and capture by overly general "black-hole skills". Execution law: before state realization, joint routing is approximately multiplicative, whereas correct execution can improve difficult downstream decisions by about $4{\times}$. A single parameter, the routing logarithmic decay slope $b$, couples the two laws: routing-side fits predict execution-side rescue across models, showing that the same library property controls both pre-execution collapse and downstream recoverability. The laws are actionable: law-guided optimization raises held-out routing accuracy from 71.3% to 91.7%, reduces hijack from 22.4% to 4.1%, and transfers directionally to downstream ClawBench and ClawMark execution settings, improving mean pass rate from 49.3% to 61.6% on ClawBench and from 28.4% to 34.5% on ClawMark. These results show that
agent performance depends not only on model capability, but also on the structure, granularity, and exposure policy of the skill library.
Read the full abstract ▾
All authors Charles Chen, Qiming Yu, Yuhang Gu, Zhuoye Huang, Hanjing Li, Hongyu Liu, Simin Liu, Jinhao Liu, Dengyun Peng, Jiangyi Wang, Zheng Yan, Fanqing Meng, Ethan Qin, Carl Che, Mengkang Hu
Computation and Language · Cross-listed from cs.DL
Generative Artificial Intelligence for Literature Reviews
Gerit Wagner, Julian Prester, Reza Mousavi et al. (+2 more)
Generative artificial intelligence (GenAI), based on large-
language models (LLMs), such as ChatGPT, has taken organizations, academia, and the public by storm. In particular, impressive GenAI capabilities such as summarization of large text corpora, question-answering, data extraction, and translation, carry profound implications for the conduct of literature reviews. This impacts science, organizations and the general public, as all can benefit from GenAI-supported literature reviews. Building on the technical foundations of GenAI and grounded in established methodological discourse, this work outlines approaches for conducting literature reviews using both general-purpose (e.g., ChatGPT, Gemini, Claude) and specialized GenAI tools (e.g., Consensus, Elicit). We provide illustrative examples of
prompts and suggest methodologically-sound literature review strategies. Throughout this perspective paper, we adopt a balanced approach considering both the opportunities and the risks of relying on GenAI in the conduct of literature reviews. We conclude by discussing philosophical questions related to the effects of GenAI on long-term scientific progress, and also present fruitful opportunities for research on improving the core of GenAI's technology-its architecture and training data-and suggest open issues in GenAI-based literature reviews methodology.
Read the full abstract ▾
All authors Gerit Wagner, Julian Prester, Reza Mousavi, Roman Lukyanenko, Guy Pare
Computation and Language · Cross-listed from cs.CV
Mechanistically Interpretable Neural Encoding Reveals Fine-Grained Functional Selectivity in Human Visual Cortex
Idan Daniel Grosbard, Mor Geva and Galit Yovel
A central goal in understanding human
vision is to uncover the visual features that drive neuronal activity. A growing body of work has used artificial neural networks as encoding models to predict cortical responses to natural images, revealing the visual content that activates category-selective regions. However, existing approaches are largely correlational and treat the encoder as a black box, leaving open which image features drive each voxel's response. We introduce Mechanistically Interpretable Neural Encoding (MINE), a framework that opens this black box by applying mechanistic-interpretability tools to localize the features within natural images that drive millimeter-scale (voxel-level) activity. MINE predicts each voxel's response using
language-aligned image representations, and produces semantically interpretable descriptions of the features critical for the voxel's activation. We further generalize these per-image features into per-voxel functional profiles. To validate the per-image descriptions, we show they are sufficient to generate images that elicit voxel responses matching the responses to the original images, more accurately than images generated from random or low-attribution controls. Moreover, counterfactually inserting or removing the predicted features from images shifts activation in the expected direction, providing causal evidence. Counterfactual editing guided by the per-voxel activation profiles produces even stronger activation shifts, indicating that the profiles faithfully capture each voxel's selectivity. Finally, we apply MINE to well-studied category-selective
brain regions, showing it recovers their known categorical preferences while revealing fine-grained unique voxel structure within each region. Overall, our results establish mechanistic interpretability as a path to discover and causally validate fine-grained hypotheses about neural function.
Read the full abstract ▾
Computation and Language
A Generative AI Framework for Intelligent Utility Billing CO 2 Analytics and Sustainable Resource Optimisation
Pavan Manjunath and Thomas Pruefer
Distribution utilities are now expected to deliver bills that customers can actually read attach a defensible carbon number to every kWh sold and schedule load against grid stress and emissions constraints We propose an end-to-end framework that unifies four production-grade capabilities under one architectural roof a generative-AI
agent that drafts each customers natural-
language billing statement from structured numeric inputs under a constrained decoding policy a
transformer-based forecaster that supplies the day-ahead consumption estimate with calibrated
quantile bands
Read the full abstract ▾
Computation and Language · Cross-listed from cs.CY
AI-Mediated Communication Can Steer Collective Opinion
Stratis Tsirtsis, Kai Rawal, Chris Russell et al. (+2 more)
Generative artificial intelligence (AI) is increasingly integrated into the online platforms where humans exchange opinions; large
language models (LLMs) now polish users' posts on LinkedIn and provide context for content shared on X. While prior work has shown that AI can express biased opinions and shape individuals' opinions during human-AI interactions, less
attention has been paid to its influence on collective opinion formation when mediating human-to-human communication. We address this gap via a combination of empirical and theoretical analyses. We show empirically that LLMs from multiple popular families introduce directional biases when instructed to edit human-written texts on contested topics, for example, nudging texts in favor of gun control and against atheism. Building on this observation, we introduce a mathematical model of opinion dynamics in which an AI system sits between users on a social network, transforming the opinions they express and perceive. By analytically characterizing the
equilibrium of this model and performing simulations on real social network data, we show that biases introduced by AI in human-to-human communication can be amplified through the network and shift collective opinion in their direction. In light of these findings, we investigate whether such biases are controllable by online platforms. We audit the "Explain this post" feature on X and find evidence of pro-life bias in Grok's outputs on abortion-related content, which we trace back to specific design choices. We conclude with a discussion of the broader implications of our findings in relation to ongoing legislative efforts in the European Union.
Read the full abstract ▾
All authors Stratis Tsirtsis, Kai Rawal, Chris Russell, Brent Mittelstadt, Sandra Wachter
Computation and Language · Cross-listed from cs.LG
No Free Swap: Protocol-Dependent Layer Redundancy in Transformers
Gabriel Garcia
When researchers ask whether two
transformer layers are "equivalent" for compression, they often conflate distinct tests. Replacement asks whether one layer's map can substitute for another's in place; interchange asks whether two layers approximately commute when their positions are swapped. Both are output-grounded swap-KL probes, but they need not agree: on pretrained
transformers the protocol gap can change which layers look safe to prune by several-fold under the same evaluator, especially when replacement distances are high. We measure both protocols across checkpoints and architectures. On a Pythia training trajectory (410M and 1.4B), the replacement-interchange gap grows from initialization to convergence. Under one matched WikiText-2 contract at 8B scale, Qwen3-8B enters a divergent regime: interchange-guided removal is several-fold safer than replacement-guided at the same layer budgets, while Llama-3.1-8B ties the two protocols for
pruning cost even though interchange KL is lower, showing metric gaps need not map one-to-one to removal. Before layer removal or merging, score both swap-KLs on the target checkpoint; the diagnostic requires only unlabeled forward passes.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.AI
FORGE: Self-Evolving Agent Memory With No Weight Updates via Population Broadcast
Igor Bogdanov, Chung-Horng Lung, Thomas Kunz et al. (+3 more)
Can LLM
agents improve decision-making through self-generated
memory without gradient updates? We propose FORGE (Failure-Optimized Reflective Graduation and
Evolution), a staged, population-based protocol that evolves
prompt-injected natural-
language memory for hierarchical ReAct
agents. FORGE wraps a Reflexion-style inner loop, where a dedicated reflection
agent (using the same underlying LLM, no
distillation from a stronger model) converts failed trajectories into reusable knowledge artifacts: textual heuristics (Rules), few-shot demonstrations (Examples), or both (Mixed), with an outer loop that propagates the best-performing instance's
memory to the population between stages and freezes converged instances via a graduation criterion. We evaluate on CybORG CAGE-2, a
stochastic network-defense POMDP at a 30-step horizon against the B-line attacker, where all four tested LLM families (Gemini-2.5-Flash-Lite, Grok-4-Fast, Llama-4-Maverick, Qwen3-235B) exhibit strongly negative, heavy-tailed zero-shot rewards. Compared against both a zero-shot baseline and a Reflexion baseline (isolated single-stream learning), FORGE improves average evaluation return by 1.7-7.7$\times$ over zero-shot and by 29-72% over Reflexion in all 12 model-representation conditions, reducing major-failure rates (below $-100$) to as low as $\sim$1%. We find that (1) population broadcast is critical mechanism, with a no-graduation ablation confirming that broadcast carries the performance gains while graduation primarily saves compute; (2) Examples achieves the strongest returns for three of four models, Rules offers the best cost-reliability profile with $\sim$40% fewer
tokens; and (3) weaker baseline models benefit disproportionately, suggesting FORGE may mitigate capability gaps rather than amplify strong models. All evidence is confined to CAGE-2 B-line; cross-family findings are directional evidence.
Read the full abstract ▾
All authors Igor Bogdanov, Chung-Horng Lung, Thomas Kunz, Jie Gao, Adrian Taylor, Marzia Zaman
Computation and Language
A Unified Generative-AI Framework for Smart Energy Infrastructure: Intelligent Gas Distribution, Utility Billing, Carbon Analytics, and Quantum-Inspired Optimisation
Pavan Manjunath and Thomas pruefer
The accelerating convergence of smart metering, generative artificial intelligence, and
quantum-inspired combinatorial optimisation is reshaping how energy utilities manage physical infrastructure, customer engagement, and environmental accountability
Computation and Language
Artificial Aphasias in Lesioned Language Models
Nathan Roll, Jill Kries, Laura Gwilliams et al. (+1 more)
Aphasias, selective
language impairments which can arise from
brain damage, reveal the functional organization of human
language by providing causal links between affected
brain regions and specific symptom profiles. Drawing on this literature, we introduce an aphasia-inspired technique to characterize the emergent functional organization of
language models (LMs). We “lesion” (zero-out) model parameters and measure the effects of this intervention against clinical aphasia symptoms, as diagnosed by the Text Aphasia Battery (TAB). When applied to 112,426 outputs from five 1B-scale LMs, the full range of evaluated symptoms surface, but in distributions largely distinct from those of humans. Our method uncovers broad symptom-profile differences between
attention components (query, key, value, output) and feed-forward components (up, gate, down), with weaker evidence for differences among components within the same mechanism. We also find an effect of depth, where lesions in early layers disproportionately cause syntactic and semantic symptoms while late-middle layers yield higher rates of phonological and fluency deficits. Although some LM lesions induce quantitatively more similar profiles to some human aphasia types than others, qualitative differences in symptom patterns between LMs and humans suggest that aphasia syndromes are heavily influenced by the details of learning and processing rather than being a domain-invariant consequence of disrupted
language processing.
Read the full abstract ▾
All authors Nathan Roll, Jill Kries, Laura Gwilliams, Cory Shain
Computation and Language
Argus: Evidence Assembly for Scalable Deep Research Agents
Zhen Zhang, Liangcai Su, Zhuo Chen et al. (+7 more)
Deep research
agents have achieved remarkable progress on complex information seeking tasks. Even long ReAct style rollouts explore only a single trajectory, while recent state of the art systems scale inference time compute via parallel search and aggregation. Yet deep research answers are composed of complementary pieces of evidence, which parallel rollouts often duplicate rather than complete, yielding diminishing returns while pushing the aggregation context toward the model's limit. We propose Argus, an agentic system in which a Searcher and a Navigator cooperate to treat deep research as assembling a jigsaw from complementary evidence pieces, rather than brute forcing the whole answer in parallel. The Searcher collects evidence traces for a given sub-query through ReAct-style interaction. The Navigator maintains a shared evidence graph, verifying which pieces are still missing, dispatching Searchers to gather them, and
reasoning over the completed graph to produce a source-traced final answer. We train the Navigator with reinforcement learning to verify, dispatch, and synthesize, while independently training the Searcher to remain a standard ReAct
agent. The resulting Navigator supports rollouts with a single Searcher or many in parallel without retraining. With both Searcher and Navigator built on a 35B-A3B MoE backbone, Argus gains 5.5 points with a single Searcher and 12.7 points with 8 parallel Searchers, averaged over eight
benchmarks. With 64 Searchers it reaches 86.2 on BrowseComp, surpassing every proprietary
agent we
benchmark, while the Navigator's
reasoning context stays under 21.5K
tokens.
Read the full abstract ▾
All authors Zhen Zhang, Liangcai Su, Zhuo Chen, Xiang Lin, Haotian Xu, Simon Shaolei Du, Kaiyu Yang, Bo An, Lidong Bing, Xinyu Wang
Computation and Language · Cross-listed from cs.AI
Fully Open Meditron: An Auditable Pipeline for Clinical LLMs
Xavier Theimer-Lienhard, Mushtaha El-Amin, Fay Elhassan et al. (+5 more)
Clinical decision support systems (CDSS) require scrutable, auditable pipelines that enable rigorous, reproducible validation. Yet current LLM-based CDSS remain largely opaque. Most "open" models are open-weight only, releasing parameters while withholding the data provenance, curation procedures, and generation pipelines that determine model behavior. Fully Open (FO) models, which expose the complete training stack end-to-end, do not currently exist in medicine. We introduce Fully Open Meditron, the first fully open pipeline for building LLM-CDSS, comprising a clinician-audited training corpus, a reproducible data construction and training framework, and a use-aligned evaluation protocol. The corpus unifies eight public medical QA
datasets into a normalized conversational format and expands coverage with three clinician-vetted synthetic extensions: exam-style QA, guideline-grounded QA derived from 46,469 clinical practice guidelines, and clinical vignettes. The pipeline enforces system-wide decontamination, gold-label resampling of teacher generations, and end-to-end validation by a four-physician panel. We evaluate using an LLM-as-a-judge protocol over expert-written clinical vignettes, calibrated against 204 human raters. We apply the recipe to five FO base models (Apertus-70B/8B-Instruct, OLMo-2-32B-SFT, EuroLLM-22B/9B-Instruct). All MeditronFO variants are preferred over their bases. Apertus-70B-MeditronFO improves +6.6 points over its base (47.2% to 53.8%) on aggregate medical
benchmarks, establishing a new FO SoTA. Gemma-3-27B-MeditronFO is preferred over MedGemma in 58.6% of LLM-as-a-judge comparisons and outperforms it on HealthBench (58% vs 55.9%). These results show that fully open pipelines can achieve state-of-the-art domain-specific performance without sacrificing auditability or reproducibility.
Read the full abstract ▾
All authors Xavier Theimer-Lienhard, Mushtaha El-Amin, Fay Elhassan, Sahaj Vaidya, Victor Cartier-Negadi, David Sasu, Lars Klein, Mary-Anne Hartley
Computation and Language · Cross-listed from cs.AI
Confirming Correct, Missing the Rest: LLM Tutoring Agents Struggle Where Feedback Matters Most
Tahreem Yasir, Wenbo Li, Sam Gilson et al. (+3 more)
Effective tutoring requires distinguishing optimal, valid but suboptimal, and incorrect student solutions, a distinction central to intelligent tutoring systems (ITS) but untested for LLM-based tutors. As LLMs are increasingly explored as conversational complements to ITS, evaluating their diagnostic precision is essential. We present a
benchmark of seven LLM feedback
agents in propositional logic using knowledge-graph-derived ground truth across 10,836 solution–feedback pairs and three feedback conditions. Models achieved near-ceiling performance on optimal steps but systematically over-rejected valid but suboptimal
reasoning and over-validated incorrect solutions, precisely where adaptive tutoring matters most. These failures persisted across models regardless of solution context, suggesting architectural rather than informational limits. Moreover, accurate diagnosis did not reliably produce pedagogically actionable feedback, revealing a gap between diagnostic judgment and instructional effectiveness. Our findings suggest that LLMs are better suited for hybrid architectures where KG-grounded models handle diagnosis while LLMs support open-ended scaffolding and dialogue.
Read the full abstract ▾
All authors Tahreem Yasir, Wenbo Li, Sam Gilson, Sutapa Dey Tithi, Xiaoyi Tian, Tiffany Barnes
Computation and Language · Cross-listed from cs.AI
Context, Reasoning, and Hierarchy: A Cost-Performance Study of Compound LLM Agent Design in an Adversarial POMDP
Igor Bogdanov, Chung-Horng Lung, Thomas Kunz et al. (+3 more)
Deploying compound LLM
agents in
adversarial, partially observable sequential environments requires navigating several design dimensions: (1) what the
agent sees, (2) how it reasons, and (3) how tasks are decomposed across components. Yet practitioners lack guidance on which design choices improve performance versus merely increase inference costs. We present a controlled study of compound LLM
agent design in CybORG CAGE-2, a cyber defense environment modeled as a Partially Observable Markov Decision Process (POMDP). Reward is non-positive, so all configurations operate in a failure-mitigation mode. Our evaluation spans five model families, six models, and twelve configurations (3,475 episodes) with
token-level cost accounting. We vary context representation (raw observations vs. a deterministic state-tracking layer with compressed history), deliberation (self-questioning, self-critique, and self-improvement tools, with optional chain-of-thought prompting), and hierarchical decomposition (monolithic ReAct vs. delegation to specialized sub-
agents). We find that: (1) Programmatic state abstraction delivers the largest returns per
token spent (RPTS), improving mean return by up to 76% over raw observations. (2) Distributing deliberation tools across a hierarchy degrades performance relative to hierarchy alone for all five model families, reaching up to 3.4$\times$ worse mean return while using 1.8-2.7$\times$ more
tokens. We call this destructive pattern a deliberation cascade. (3) Hierarchical decomposition without deliberation achieves the best absolute performance for most models, and context engineering is generally more cost-effective than deliberation. These findings suggest a design principle for structured
adversarial POMDPs: invest in programmatic infrastructure and clean task decomposition rather than deeper per-
agent reasoning, as these strategies can interfere when combined.
Read the full abstract ▾
All authors Igor Bogdanov, Chung-Horng Lung, Thomas Kunz, Jie Gao, Adrian Taylor, Marzia Zaman
Computation and Language
Improving Cross-Cultural Survey Simulation with Calibrated Value Personas
Axel Abels, Elias Fernandez Domingos, Apurva Shah et al. (+1 more)
Large
language models (LLMs) are increasingly used to simulate human opinions and survey responses, but their ability to reproduce population responses across cultures remains limited. Existing persona-based prompting methods typically rely on sociodemographic or personality traits, which are only indirect proxies for the values that shape human responses. We propose a value-based persona construction method that derives textual descriptors from survey responses capturing core cultural dimensions. By sampling value profiles from target populations and aggregating LLM responses across personas, we obtain population-level predictions grounded in observed value distributions. We further introduce a calibration procedure that improves response diversity while preserving estimated opinions. We show that our approach reduces prediction error across countries, with the largest improvements observed in underrepresented populations. This substantially narrows the performance gap between countries aligned with dominant LLM priors and those that are less represented in training data, while also yielding response distributions that closely match human diversity.
Read the full abstract ▾
All authors Axel Abels, Elias Fernandez Domingos, Apurva Shah, Tom Lenaerts
Computation and Language
Optimized Three-Dimensional Photovoltaic Structures with LLM guided Tree Search
Michael P. Brenner, Lizzie Dorfman and John C. Platt
We present a case study for how AI coding systems can be used to generate novel scientific hypotheses. We combine a generic coding
agent (Google's AntiGravity) with an LLM-driven tree search
algorithm (Empirical Research Assistance / ERA) to autonomously generate high-efficiency three-dimensional photovoltaic (3DPV) structures that overcome losses limiting flat
solar panels at mid-latitudes. These structures operate by presenting favorable angles to the sun throughout the day, and for illustrative purposes we focus on optimizing performance for a single
solar day. Our workflow begins by using AntiGravity to reproduce calculations \cite{bernardi2012
solar} showing that 3DPV can have energy densities much higher than stationary flat PV panels. We use these initial designs as the starting point for large scale tree search, where we seek improved solutions and score them for their diurnal yield. The initial tree search leads to nominally more efficient solutions, yet they are caused by algorithmic reward hacking, arising from non-physical design features such as structurally levitating disconnected tiers and exploitations of the discretizations in the optics solver. To counteract this, we develop a workflow where the coding
agent iteratively patches the physics engine with constraints to eliminate reward hacking. With reward-hacking eliminated, ERA discovers a series of designs with various constraints and improved performance, including optimal designs with different fixed collector areas, optimizing zenith tracking and avoiding self shadowing. Combining coding
agents with tree search (ERA) provides a powerful platform for scientific discovery, for problems whose solutions can be empirically evaluated with a score function.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.AI
Look Before You Leap: Autonomous Exploration for LLM Agents
Ziang Ye, Wentao Shi, Yuxin Liu et al. (+6 more)
Large
language model based
agents often fail in unfamiliar environments due to premature exploitation: a tendency to act on prior knowledge before acquiring sufficient environment-specific information. We identify autonomous exploration as a critical yet underexplored capability for building adaptive
agents. To formalize and quantify this capability, we introduce Exploration Checkpoint Coverage, a verifiable metric that measures how broadly an
agent discovers key states, objects, and affordances. Our systematic evaluation reveals that
agents trained with standard task-oriented reinforcement learning consistently exhibit narrow and repetitive behaviors that impede downstream performance. To address this limitation, we develop a training strategy that interleaves task-execution rollouts and exploration rollouts, with each type of rollout optimized by its corresponding verifiable reward. Building on this training strategy, we propose the Explore-then-Act paradigm, which decouples information-gathering from task execution:
agents first utilize an interaction budget to acquire grounded environmental knowledge, then leverage it for task resolution. Our results demonstrate that learning to systematically explore is imperative for building generalizable and real-world-ready
agents.
Read the full abstract ▾
All authors Ziang Ye, Wentao Shi, Yuxin Liu, Yu Wang, Zhengzhou Cai, Yaorui Shi, Qi Gu, Xunliang Cai, Fuli Feng
Computation and Language
SGR: A Stepwise Reasoning Framework for LLMs with External Subgraph Generation
Xin Zhang, Yang Cao, Baoxing Wu et al. (+2 more)
Large
Language Models (LLMs) have demonstrated strong capabilities across diverse NLP applications, such as translation, text generation, and question answering. Nevertheless, they remain limited in complex settings that demand deep
reasoning and logical inference. Since these models are trained on large-scale text corpora, their generation process may still introduce irrelevant, noisy, or factually inconsistent content. To mitigate this problem, we introduce SGR, a stepwise framework that enhances LLM
reasoning through external subgraph generation. SGR builds query-specific subgraphs from external knowledge bases and uses their semantic structure to support multi-step inference. By grounding intermediate
reasoning steps in structured external knowledge, the framework helps the model concentrate on relevant entities, relations, and supporting evidence. In particular, SGR first constructs a subgraph tailored to the input question. It then guides the model to reason progressively over the generated structure and combines multiple
reasoning trajectories to obtain the final prediction. Experimental results across several
benchmark datasets show that SGR achieves consistent improvements over competitive baselines, highlighting its value for improving both
reasoning accuracy and factual reliability.
Read the full abstract ▾
All authors Xin Zhang, Yang Cao, Baoxing Wu, Kai Song, Siying Li
Computation and Language
DebiasRAG: A Tuning-Free Path to Fair Generation in Large Language Models through Retrieval-Augmented Generation
Rui Chu, Bingyin Zhao, Thanh Quoc Hung Le et al. (+6 more)
Large
language models (LLMs) have achieved unprecedented success due to their exceptional generative capabilities. However, because they depend on knowledge encapsulated from training corpora, they may produce
hallucinations, stereotypes, and socially biased content. In particular, LLMs are prone to prejudiced responses involving race, gender, and age, which are collectively referred to as social biases. Prior studies have used fine-tuning and
prompt engineering to mitigate such biases in LLMs, but these methods require additional training resources or domain knowledge to design the framework. Moreover, they may degrade the original capabilities of LLMs and often overlook the need for dynamic debiasing contexts for fairer inference. In this paper, we propose DebiasRAG, a novel tuning-free and dynamic query-specific debiasing framework based on
retrieval-augmented generation (RAG). DebiasRAG improves
fairness while preserving the intrinsic properties of LLMs, such as representation ability. DebiasRAG consists of three stages: (1) query-specific debiasing candidate generation; (2) context candidate pool construction; and (3) gradient-updated debiasing-guided context piece reranking. First, DebiasRAG leverages self-diagnosed bias contexts relevant to the query through regular
retrieval, where the bias contexts are prepared offline by the DebiasRAG provider. Given the query-specific bias contexts, DebiasRAG reversely produces debiasing contexts, which are provided as additional
fairness constraints for LLM outputs. Second, a regular RAG
retrieval process produces query-related contexts from the regular RAG document database, such as a chunked Wikipedia
dataset.
Read the full abstract ▾
All authors Rui Chu, Bingyin Zhao, Thanh Quoc Hung Le, Duy Cao Hoang, Huawei Lin, Ping Li, Weijie Zhao, Khoa D Doan, Yingjie Lao
Computation and Language
Multi-Level Contextual Token Relation Modeling for Machine-Generated Text Detection
Chenwang Wu, Yiuming Cheung, Bo Han et al. (+2 more)
Machine-generated texts (MGTs) pose risks such as disinformation and phishing, underscoring the need for reliable detection. Metric-based methods, which extract statistically distinguishable features of MGTs, are often more practical than complex model-based methods that are prone to overfitting. Given their diverse designs, we first place representative metric-based methods within a unified framework, enabling a clear assessment of their advantages and limitations. Our analysis identifies a core challenge across these methods: the
token-level detection score is easily biased by the inherent randomness of the MGTs generation process. Then, we theoretically derive the multi-hop transitions of the
token-level detection score and explore their local and global relations. Based on these findings, we propose a multi-level contextual
token relation modeling framework for MGT detection. Specifically, for local relations, we model them through a lightweight Markov-informed calibration module that refines
token-level evidence before aggregation. For global relations, we introduce a rule-support
reasoning module that uses explicit logical rules derived from contextual score statistics. Finally, we combine the local calibrated score and the global rule-support
reasoning signal in a joint multi-level inference framework. Extensive experiments show broad and substantial improvements across various real-world scenarios, including cross-LLM and cross-domain settings, with low computational overhead.
Read the full abstract ▾
All authors Chenwang Wu, Yiuming Cheung, Bo Han, Shuhai Zhang, Defu Lian
Computation and Language
Can Large Language Models Imitate Human Speech for Clinical Assessment? LLM-Driven Data Augmentation for Cognitive Score Prediction
Si-Belkacem Yamine Ketir, Lenard Paulo Tamayo, Shohei Hisada et al. (+3 more)
Accurate assessment of cognitive decline from spontaneous
speech remains challenging due to limited
dataset size and class imbalance. In this work, we propose a large
language model (LLM)-driven data augmentation framework to improve the prediction of cognitive scores from
speech. Experiments are conducted on a Japanese corpus in which each participant provides both a spontaneous oral narrative and a written response to the same clinical
prompt. The written responses serve as semantic anchors to generate multiple oral-like monologues in different styles using GPT-5. We then predict Hasegawa Dementia Scale scores, a widely used cognitive screening tool in Japan, using a Partial Least Squares regression model trained on Sentence-BERT
speech embeddings. We investigate two augmentation strategies: random class-balanced selection, which yields moderate but unstable improvements, and similarity-guided class-balanced selection. The latter prioritizes semantically close synthetic samples, leading to more consistent improvements and substantially reducing prediction error for minority low-score participants while maintaining performance for the majority group. Overall, our findings demonstrate the potential of semantically guided LLM-driven augmentation as a principled approach for addressing class imbalance and improving data efficiency in clinical
speech analysis.
Read the full abstract ▾
All authors Si-Belkacem Yamine Ketir, Lenard Paulo Tamayo, Shohei Hisada, Shaowen Peng, Shoko Wakamiya, Eiji Aramaki
Computation and Language · Cross-listed from cs.AI
Reasoners or Translators? Contamination-aware Evaluation and Neuro-Symbolic Robustness in Tax Law
Parisa Kordjamshidi, Samer Aslan, Madhavan Seshadri et al. (+2 more)
Recent advances in large
language models (LLMs) have significantly enhanced automated legal
reasoning. Yet, it remains unclear whether their performance reflects genuine legal
reasoning ability or artifacts of data contamination. We present a comprehensive empirical study of tax law
reasoning approaches and implement a contamination detection protocol to rigorously assess LLM reliability. We show that performance can be inflated by contamination. Building on this analysis, we conduct a systematic evaluation, comparing monolithic LLMs with hybrid systems that translate statutory text into formal representations and delegate inference to symbolic solvers. We build a novel test suite designed to probe generalization to unseen documents via case and rule variations. Our findings indicate that legal
reasoning is inherently compositional and that neuro-symbolic frameworks offer a more reliable and robust foundation for legal AI, as well as improved generalization to unobserved situations.
Read the full abstract ▾
All authors Parisa Kordjamshidi, Samer Aslan, Madhavan Seshadri, Leslie Barrett, Enrico Santus
Computation and Language
RecMem: Recurrence-based Memory Consolidation for Efficient and Effective Long-Running LLM Agents
Zijie Dai, Shiyuan Deng, Sheng Guan et al. (+4 more)
Memory systems often organize user-
agent interactions as retrievable external
memory and are crucial for long-running
agents by overcoming the limited context windows of LLMs. However, existing
memory systems invoke LLMs to process every incoming interaction for
memory extraction, and such an eager
memory consolidation scheme leads to substantial
token consumption. To tackle this problem, we propose RecMem by rethinking when
memory consolidation should be conducted. RecMem stores incoming interactions in a subconscious
memory layer and encode them using lightweight
embedding models for
retrieval. LLMs are only invoked to extract episodic and semantic
memory when sustained recurrence are observed for semantically similar interactions. Such recurrence-based consolidation works because these interactions correspond to a semantic cluster with rich information and thus are worth extraction and summarization. To improve accuracy, RecMem also incorporates a semantic refinement mechanism that recovers the fine-grained facts omitted by
memory extraction. Experiments show that RecMem reduces the
memory construction
token cost of three SOTA
memory systems by up to 87% while exceeding their accuracy.
Read the full abstract ▾
All authors Zijie Dai, Shiyuan Deng, Sheng Guan, Yizhou Tian, Xin Yao, Xiao Yan, James Cheng
Computation and Language
Conditioned in Training, Dispensable at Inference: Typological Conditioning for Multilingual Speech Translation
Yu Pan, Tianjiao Zhang, Yang Hou et al. (+5 more)
Multilingual speech translation frontends built on
speech large
language models typically condition on
language identity, as a learned
embedding or a prompted
language name. Such a signal, however, is learned separately for each
language and transfers nothing between typologically related
languages. In this paper, we present TypoCast, a frontend that conditions on the typological profile of the source
language rather than on its label alone. A typology-informed hierarchical
language encoding represents each
language by its morphological type, constituent order and family, so that supervision is shared among related
languages. A dual-route mixture-of-experts adapter routes inputs through a shared typological route together with a per-
language residual, and a typology-aware
prompt conveys the same profile to the decoder. Extensive experiments on CVSS-C demonstrate that TypoCast surpasses state-of-the-art
speech-to-
speech translation systems on French, German and Spanish to English, averaged over the three directions, and outperforms the identity-conditioned S2ST-Omni on every
language of a seven-
language inventory, with only 228M trainable parameters. Ablation studies further reveal that the conditioning signal acts during training and is dispensable at inference: a model trained without it is consistently worse, whereas removing it from a trained model changes BLEU by less than 0.2, and expert usage is consistent with the signal shaping how adapter capacity is divided among
languages.
Read the full abstract ▾
All authors Yu Pan, Tianjiao Zhang, Yang Hou, Xiongfei Wu, Liang Zhang, Yves Le Traon, Lei Ma, Jianjun Zhao
Computation and Language
Judge Circuits Explain Format-Induced Inconsistency in LLM-as-a-Judge
Nils Feldhus, Tanja Baeumel, Elena Golimblevskaia et al. (+10 more)
LLM-as-a-judge has become the dominant paradigm for grading model outputs at scale, yet the same model assigns systematically different scores when its output format changes (e.g., a 1-5 rating vs. a True/False label). Existing diagnoses of these format-induced inconsistencies stop at the input-output level. Using Position-aware Edge Attribution Patching (PEAP), we causally investigate the internal mechanism in five open-weight instruction-tuned models (Gemma-3, Qwen2.5, Llama-3.1) across five judgment tasks. We find that judgments across structured understanding and open-ended preference tasks share a sparse Latent Evaluator sub-graph in the mid-to-late layers; zero-ablating it collapses judgment while damaging knowledge probes substantially less than a random ablation of equal size in architecturally modular models. By structurally decoupling abstract judging from output formatting, we provide a mechanistic account of format-induced inconsistency on the open-weight models we study: a continuous judgment signal computed in the shared trunk is mapped through fragile, format-specific terminal branches. The judgment itself can therefore be read out independently of the requested output format. Our findings imply that
benchmark comparisons of judge reliability across formats partly measure the fragile formatting stage, and can understate the quality of the underlying evaluation.
Read the full abstract ▾
All authors Nils Feldhus, Tanja Baeumel, Elena Golimblevskaia, Qianli Wang, Van Bach Nguyen, Aaron Louis Eidt, Selin Kahvecioglu, Christopher Ebert, Wojciech Samek, Jing Yang, Vera Schmitt, Sebastian Möller, Simon Ostermann