Peer Revenue Play credits. Real papers.

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Computation and Language

cs.CL Watch Computation and Language

Tuesday 6 January

Computation and Language

AfriEconQA: A Benchmark for Quantitative and Temporal Reasoning over World Bank Economic Reports

Edward Ajayi, Mustapha Alaba, David Stephen et al. (+1 more)

Reliable question answering over long institutional documents requires more than topical retrieval: a system must localize the exact passage that supports a claim and preserve precise numerical and temporal detail when the same indicator recurs across years, countries, and projection horizons in dense, repetitive prose. Existing question-answering benchmarks rarely test this combination of exact grounding and temporal precision at document scale. We introduce AfriEconQA, a benchmark for document-grounded question answering built from 204 World Bank economic reports on African economies, a corpus whose claims are tied to specific fiscal periods, countries, and projection states. AfriEconQA contains 4,138 evidence-linked QA instances across five reasoning categories: Factoid (1,063), List (852), Multiple Choice (889), Synthesis (671), and Comparison (663), targeting quantitative extraction, set recovery, discrimination, causal integration, and cross-period or cross-country reasoning. Each instance carries supporting evidence and source provenance and is constructed through an agentic generation pipeline with evidence-grounding checks and a stratified human-validation subset for gold-label auditing. We evaluate Qwen 3.6 35B, DeepSeek v4-pro, Gemma 4 12B IT, and frontier model GPT-5.5 under zero-shot, oracle, and hybrid retrieval-augmented generation (RAG) conditions on the held-out test split (n = 826). Across all four models, retrieval yields substantial gains, yet the best RAG system reaches only 0.594 F1, with residual errors concentrated in list extraction, synthesis, and temporally scoped comparison. AfriEconQA therefore poses a hard open challenge for exact numerical and temporal grounding over long institutional economic reports.
Read the full abstract

All authors Edward Ajayi, Mustapha Alaba, David Stephen, Peace Bakare

Computation and Language · Cross-listed from cs.LG

PromptPort: A Reliability Layer for Cross-Model Structured Extraction

Varun Kotte

Structured extraction with LLMs fails in production not because models lack understanding, but because output formatting is unreliable across models and prompts. A prompt that returns clean JSON on GPT-4 may produce fenced, prose-wrapped, or malformed output on Llama, causing strict parsers to reject otherwise correct extractions. We formalize this as format collapse and introduce a dual-metric evaluation framework: ROS (strict parsing, measuring operational reliability) and CSS (post-canonicalization, measuring semantic capability). On a 37,346-example camera metadata benchmark across six model families, we find severe format collapse (for example, Gemma-2B: ROS 0.116 versus CSS 0.246) and large cross-model portability gaps (0.4 to 0.6 F1). We then present PromptPort, a reliability layer combining deterministic canonicalization with a lightweight verifier (DistilBERT) and a safe-override policy. PromptPort recovers format failures (plus 6 to 8 F1), adds verifier-driven semantic selection (plus 14 to 16 F1 beyond canonicalization), and approaches per-field oracle performance (0.890 versus 0.896 in zero-shot) without modifying base models. The method generalizes to held-out model families and provides explicit abstention when uncertain, enabling reliable structured extraction in production deployments.
Read the full abstract

Computation and Language · Cross-listed from cs.SE

Sphinx: Benchmarking and Modeling for LLM-Driven Pull Request Review

Daoan Zhang, Shuo Zhang, Zijian Jin et al. (+3 more)

Pull request (PR) review is essential for ensuring software quality, yet automating this task remains challenging due to noisy supervision, limited contextual understanding, and inadequate evaluation metrics. We present Sphinx, a unified framework for LLM-based PR review that addresses these limitations through three key components: (1) a structured data generation pipeline that produces context-rich, semantically grounded review comments by comparing pseudo-modified and merged code; (2) a checklist-based evaluation benchmark that assesses review quality based on structured coverage of actionable verification points, moving beyond surface-level metrics like BLEU; and (3) Checklist Reward Policy Optimization (CRPO), a novel training paradigm that uses rule-based, interpretable rewards to align model behavior with real-world review practices. Extensive experiments show that models trained with Sphinx achieve state-of-the-art performance on review completeness and precision, outperforming both proprietary and open-source baselines by up to 40% in checklist coverage. Together, Sphinx enables the development of PR review models that are not only fluent but also context-aware, technically precise, and practically deployable in real-world development workflows. The data will be released after review.
Read the full abstract

All authors Daoan Zhang, Shuo Zhang, Zijian Jin, Jiebo Luo, Shengyu Fu, Elsie Nallipogu

Computation and Language

EpiQAL: Benchmarking Large Language Models in Epidemiological Question Answering and Reasoning

Mingyang Wei, Dehai Min, Zewen Liu et al. (+8 more)

Reliable epidemiological reasoning requires synthesizing study evidence to infer disease burden, transmission dynamics, and intervention effects at the population level. Existing medical question answering benchmarks primarily emphasize clinical knowledge or patient-level reasoning, yet few systematically evaluate evidence-grounded epidemiological inference. We present EpiQAL, to our knowledge the first diagnostic benchmark for epidemiological question answering over research literature, comprising three subsets built from open-access articles across diverse diseases. The three subsets progressively test factual recall, multi-step inference, and conclusion reconstruction under incomplete information, and are constructed through a quality-controlled pipeline combining taxonomy guidance, multi-model verification, and difficulty screening. Experiments on fifteen models spanning open-source and proprietary systems reveal that current LLMs show limited performance on epidemiological reasoning, with multi-step inference posing the greatest challenge. Model rankings shift across subsets, and scale alone does not predict success. Chain-of-Thought prompting benefits multi-step inference but yields mixed results elsewhere. EpiQAL provides fine-grained diagnostic signals for evidence-grounding, inferential reasoning, and conclusion reconstruction.
Read the full abstract

All authors Mingyang Wei, Dehai Min, Zewen Liu, Yuzhang Xie, Guanchen Wu, Ziyang Zhang, Carl Yang, Max S. Y. Lau, Qi He, Lu Cheng, Wei Jin

Computation and Language · Cross-listed from econ.EM

Content vs. Form: What Drives the Writing Score Gap Across Socioeconomic Backgrounds? A Generated Panel Approach

Nadav Kunievsky and Pedro Pertusi

Students from different socioeconomic backgrounds exhibit persistent gaps in test scores, gaps that can translate into unequal educational and labor-market outcomes later in life. In many assessments, performance reflects not only what students know, but also how effectively they can communicate that knowledge. This distinction is especially salient in writing assessments, where scores jointly reward the substance of students' ideas and the way those ideas are expressed. As a result, observed score gaps may conflate differences in underlying content with differences in expressive skill. A central question, therefore, is how much of the socioeconomic-status (SES) gap in scores is driven by differences in what students say versus how they say it. We study this question using a large corpus of persuasive essays written by U.S. middle- and high-school students. We introduce a new measurement strategy that separates content from style by leveraging large language models to generate multiple stylistic variants of each essay. These rewrites preserve the underlying arguments while systematically altering surface expression, creating a "generated panel" that introduces controlled within-essay variation in style. This approach allows us to decompose SES gaps in writing scores into contributions from content and style. We find an SES gap of 0.67 points on a 1-6 scale. Approximately 69% of the gap is attributable to differences in essay content quality, Style differences account for 26% of the gap, and differences in evaluation standards across SES groups account for the remaining 5%. These patterns seems stable across demographic subgroups and writing tasks. More broadly, our approach shows how large language models can be used to generate controlled variation in observational data, enabling researchers to isolate and quantify the contributions of otherwise entangled factors.
Read the full abstract

Computation and Language

Prompting Underestimates LLM Capability for Time Series Classification

Dan Schumacher, Erfan Nourbakhsh, Rocky Slavin et al. (+1 more)

Prompt-based evaluations suggest that large language models (LLMs) perform poorly on time series classification, raising doubts about whether they encode meaningful temporal structure. We show that this conclusion reflects limitations of prompt-based generation rather than the model's representational capacity by directly comparing prompt outputs with linear probes over the same internal representations. While zero-shot prompting performs near chance, linear probes improve average F1 from 0.15-0.26 to 0.61-0.67, often matching or exceeding specialized time series models. Layer-wise analyses further show that class-discriminative time series information emerges in early transformer layers and is amplified by visual and multimodal inputs. Together, these results demonstrate a systematic mismatch between what LLMs internally represent and what prompt-based evaluation reveals, leading current evaluations to underestimate their time series understanding.
Read the full abstract

All authors Dan Schumacher, Erfan Nourbakhsh, Rocky Slavin, Anthony Rios

Computation and Language

Enhancing Linguistic Competence of Language Models through Pre-training with Language Learning Tasks

Atsuki Yamaguchi, Maggie Mi and Nikolaos Aletras

Language models (LMs) are pre-trained on raw text datasets to generate text sequences token-by-token. While this approach facilitates the learning of world knowledge and reasoning, it does not explicitly optimize for linguistic competence. To bridge this gap, we propose L2T, a pre-training framework integrating Language Learning Tasks alongside standard next-token prediction. Inspired by human language acquisition, L2T transforms raw text into structured input-output pairs to provide explicit linguistic stimulation. Pre-training LMs on a mixture of raw text and L2T data not only improves overall performance on linguistic competence benchmarks but accelerates its acquisition, while maintaining competitive performance on general reasoning tasks.
Read the full abstract

Computation and Language

Grading Scale Impact on LLM-as-a-Judge: Human-LLM Alignment Is Highest on 0-5 Grading Scale

Weiyue Li, Minda Zhao, Weixuan Dong et al. (+12 more)

Large language models (LLMs) are increasingly used as automated evaluators, yet prior works demonstrate that these LLM judges often lack consistency in scoring when the prompt is altered. However, the effect of the grading scale itself remains underexplored. We study the LLM-as-a-judge problem by comparing two kinds of raters: humans and LLMs. We collect ratings from both groups on three scales and across six benchmarks that include objective, open-ended subjective, and mixed tasks. Using intraclass correlation coefficients (ICC) to measure absolute agreement, we find that LLM judgments are not perfectly consistent across scales on subjective benchmarks, and that the choice of scale substantially shifts human-LLM agreement, even when within-group panel reliability is high. Aggregated over tasks, the grading scale of 0-5 yields the strongest human-LLM alignment. We further demonstrate that pooled reliability can mask benchmark heterogeneity and reveal systematic subgroup differences in alignment across gender groups, strengthening the importance of scale design and sub-level diagnostics as essential components of LLM-as-a-judge protocols.
Read the full abstract

All authors Weiyue Li, Minda Zhao, Weixuan Dong, Jiahui Cai, Yuze Wei, Michael Pocress, Yi Li, Wanyan Yuan, Xiaoyue Wang, Ruoyu Hou, Kaiyuan Lou, Wenqi Zeng, Yutong Yang, Yilun Du, Mengyu Wang

Computation and Language

The Critical Role of Aspects in Measuring Document Similarity

Eftekhar Hossain, Tarnika Hazra, Ahatesham Bhuiyan et al. (+1 more)

We introduce ASPECTSIM, a simple and interpretable framework that requires conditioning document similarity on an explicitly specified aspect, which is different from the traditional holistic approach in measuring document similarity. Experimenting with a newly constructed benchmark of 26K aspect-document pairs, we found that ASPECTSIM, when implemented with direct GPT-4o prompting, achieves substantially higher human-machine agreement ($\approx$80% higher) than the same for holistic similarity without explicit aspects. These findings underscore the importance of explicitly accounting for aspects when measuring document similarity and highlight the need to revise standard practice. Next, we conducted a large-scale meta-evaluation using 16 smaller open-source LLMs and 9 embedding models with a focus on making ASPECTSIM accessible and reproducible. While directly prompting LLMs to produce ASPECTSIM scores turned out be ineffective (20-30% human-machine agreement), a simple two-stage refinement improved their agreement by $\approx$140%. Nevertheless, agreement remains well below that of GPT-4o-based models, indicating that smaller open-source LLMs still lag behind large proprietary models in capturing aspect-conditioned similarity.
Read the full abstract

All authors Eftekhar Hossain, Tarnika Hazra, Ahatesham Bhuiyan, Santu Karmaker

Computation and Language · Cross-listed from cs.LG

Spectral Archaeology: The Causal Topology of Model Evolution

Valentin Noël

Behavioral benchmarks tell us what a model does, but not how. We introduce a training-free mechanistic probe using attention-graph spectra. Treating each layer as a token graph, we compute algebraic connectivity ($\lambda_2$), smoothness, and spectral entropy. Across 12 models and 10 languages, these measures yield stable “spectral fingerprints” that expose discontinuities missed by standard evaluation. We report four results. (1) Models undergoing specific curriculum transitions (e.g., code-to-chat) show an English-only, syntax-triggered connectivity failure on non-canonical constructions, reaching $\Delta\lambda_2 \approx -0.76$. We term this scar Passive-Triggered Connectivity Collapse (PTCC). Analysis of the Phi lineage reveals that PTCC appears and resolves across developmental stages, implicating brittle curriculum shifts rather than synthetic data per se. (2) PTCC reflects a specialization trade-off: strengthened formal routing at the expense of stylistic flexibility. (3) We identify four recurrent processing strategies; simple frozen-threshold rules enable perfect forensic identification across lineages. (4) Mechanistically, PTCC localizes to a sparse Layer 2 “compensatory patch” of heads that fails under syntactic stress; activation steering can partially restore connectivity, recovering $\approx 38\%$ of lost information flow. Finally, dominant topological regimes track tokenization density more than language identity, suggesting “healthy” geometry varies systematically across scripts. Overall, attention-graph spectra provide a practical tool for auditing and training-regime verification.
Read the full abstract

Computation and Language

Training-Free Adaptation of New-Generation LLMs using Legacy Clinical Models

Sasha Ronaghi, Chloe Stanwyck, Asad Aali et al. (+4 more)

Adapting language models to the clinical domain through continued pretraining and instruction tuning requires costly retraining for each new model generation. We propose Cross-Architecture Proxy Tuning (CAPT), a model-ensembling approach that enables training-free adaptation of state-of-the-art general-domain models using existing clinical models. CAPT supports models with disjoint vocabularies, leveraging contrastive decoding to selectively inject clinically relevant signals while preserving the general-domain model's reasoning and fluency. On six clinical classification and text-generation tasks, CAPT with a new-generation general-domain model and an older-generation clinical model consistently outperforms both models individually and state-of-the-art ensembling approaches (average +17.6% over UniTE, +41.4% over proxy tuning across tasks). Through token-level analysis and physician case studies, we demonstrate that CAPT amplifies clinically actionable language, reduces context errors, and increases clinical specificity. This technique especially benefits healthcare institutions with constrained computational capacity that cannot support iterative clinical training and want to adopt emerging general-domain model advances.
Read the full abstract

All authors Sasha Ronaghi, Chloe Stanwyck, Asad Aali, Amir Ronaghi, Miguel Fuentes, Tina Hernandez-Boussard, Emily Alsentzer

Computation and Language

MAPLE: Medical Aspect-Based Summarization with Phrase-Level Evidence

Bohao Chu, Hendrik Damm, Tabea M. G. Pakull et al. (+5 more)

Trustworthy clinical summarization requires every claim to be traceable to its evidence, yet existing attribution often resolves only to the sentence or document, leaving clinicians to scan surrounding text for the few words that matter. We argue that the unit of attribution should match the unit of verification: the precise phrase the reader's eye must land on. We present MAPLE (Medical Aspect-Based Summarization with Phrase-Level Evidence), a human-annotated benchmark that grounds each summarized claim in both cited sentences and contributory phrases within them. Spanning 152 randomized controlled trial (RCT) abstracts and 16 clinically motivated aspects, MAPLE comprises 1,799 aspect-based summaries with two-level evidence. We further introduce a decoupled evaluation framework that separately scores content, traceability, and locatability, together with a proxy for the amount of source text a clinician must inspect to verify a claim. Benchmarking eleven LLMs shows that sentence-level citation is consistently strong (C-F1 up to 90.9%), while phrase-level grounding remains less stable and the most discriminative axis across models (P-F1 66.1-84.5%). These results suggest that the key challenge is not only producing accurate summaries, but localizing their supporting evidence precisely enough for efficient clinical verification. Data and code are available at https://github.com/chubohao/maple.
Read the full abstract

All authors Bohao Chu, Hendrik Damm, Tabea M. G. Pakull, Sameh Frihat, Georg Lodde, Elisabeth Livingstone, Christoph M. Friedrich, Norbert Fuhr

Computation and Language

Implicit Graph, Explicit Retrieval: Towards Efficient and Interpretable Long-horizon Memory for Large Language Models

Xin Zhang, Kailai Yang, Hao Li et al. (+3 more)

Long-horizon applications increasingly require large language models (LLMs) to answer queries when relevant evidence is sparse and dispersed across very long contexts. Existing memory systems largely follow two paradigms: explicit structured memories offer interpretability but often become brittle under long-context overload, while latent memory mechanisms are efficient and stable yet difficult to inspect. We propose LatentGraphMem, a memory framework that combines implicit graph memory with explicit subgraph retrieval. LatentGraphMem stores a graph-structured memory in latent space for stability and efficiency, and exposes a task-specific subgraph retrieval interface that returns a compact symbolic subgraph under a fixed budget for downstream reasoning and human inspection. During training, an explicit graph view is materialized to interface with a frozen reasoner for question-answering supervision. At inference time, retrieval is performed in latent space and only the retrieved subgraph is externalized. Experiments on long-horizon benchmarks across multiple model scales show that LatentGraphMem consistently outperforms representative explicit-graph and latent-memory baselines, while enabling parameter-efficient adaptation and flexible scaling to larger reasoners without introducing large symbolic artifacts.
Read the full abstract

All authors Xin Zhang, Kailai Yang, Hao Li, Chenyue Li, Qiyu Wei, Sophia Ananiadou

Computation and Language

Tigrinya Number Verbalization: Rules, Algorithm, and Implementation

Fitsum Gaim and Issayas Tesfamariam

We present a systematic formalization of Tigrinya cardinal and ordinal number verbalization, addressing a gap in computational resources for the language. This work documents the canonical rules governing the expression of numerical values in spoken Tigrinya, including the conjunction system, scale words, and special cases for dates, times, and currency. We provide a formal algorithm for number-to-word conversion and release an open-source implementation. Evaluation of frontier large language models (LLMs) reveals significant gaps in their ability to accurately verbalize Tigrinya numbers, underscoring the need for explicit rule documentation. This work serves language modeling, speech synthesis, and accessibility applications targeting Tigrinya-speaking communities.
Read the full abstract

Computation and Language

Rendering Data Unlearnable by Exploiting LLM Alignment Mechanisms

Ruihan Zhang and Jun Sun

Large language models (LLMs) are increasingly trained on massive, heterogeneous text corpora, raising serious concerns about the unauthorised use of proprietary or personal data during model training. In this work, we address the problem of data protection against unwanted model learning in a realistic black-box setting. We propose Disclaimer Injection, a novel data-level defence that renders text unlearnable to LLMs. Rather than relying on model-side controls or explicit data removal, our approach exploits the models' own alignment mechanisms: by injecting carefully designed alignment-triggering disclaimers to prevent effective learning. Through layer-wise analysis, we find that fine-tuning on such protected data induces persistent activation of alignment-related layers, causing alignment constraints to override task learning even on common inputs. Consequently, models trained on such data exhibit substantial and systematic performance degradation compared to standard fine-tuning. Our results identify alignment behaviour as a previously unexplored lever for data protection and, to our knowledge, present the first practical method for restricting data learnability at LLM scale without requiring access to or modification of the training pipeline.
Read the full abstract

Computation and Language

Breaking the Assistant Mold: Modeling Behavioral Variation in LLM Based Procedural Character Generation

Maan Qraitem, Kate Saenko and Bryan A. Plummer

Procedural content generation has enabled vast virtual worlds through levels, maps, and quests, but large-scale character generation remains underexplored. We identify two alignment-induced biases in existing methods: a positive moral bias, where characters uniformly adopt agreeable stances (e.g. always saying lying is bad), and a helpful assistant bias, where characters invariably answer questions directly (e.g. never refusing or deflecting). While such tendencies suit instruction-following systems, they suppress dramatic tension and yield predictable characters, stemming from maximum likelihood training and assistant fine-tuning. To address this, we introduce PersonaWeaver, a framework that disentangles world-building (roles, demographics) from behavioral-building (moral stances, interactional styles), yielding characters with more diverse reactions and moral stances, as well as second-order diversity in stylistic markers like length, tone, and punctuation. Code: https://github.com/mqraitem/Persona-Weaver
Read the full abstract

Computation and Language

Metaphors are a Source of Cross-Domain Misalignment of Large Reasoning Models

Zhibo Hu, Chen Wang, Yanfeng Shu et al. (+2 more)

Earlier research has shown that metaphors influence human decision-making, raising the question of whether metaphors also influence large language models (LLMs)' reasoning pathways, given that their training data contain a large number of metaphors. In this work, we investigate the problem in the scope of the emergent misalignment problem, where LLMs can generalize patterns learned from misaligned content in one domain to another domain. We find strong evidence that metaphors in training data contribute to cross-domain misalignment in LLMs' reasoning outputs. With metaphor-based interventions during continued pre-training and fine-tuning for inducing misalignment, models exhibit significantly different degrees of emergent cross-domain misalignment. We also observe similar effects in re-alignment settings. As we further investigate this phenomenon, we find that metaphors are linked to the activation of latent features in large reasoning models. By monitoring these latent features, we design a detector that predicts misaligned content with high accuracy.
Read the full abstract

All authors Zhibo Hu, Chen Wang, Yanfeng Shu, Hye-young Paik, Liming Zhu

Computation and Language · Cross-listed from cs.CV

RiskCueBench: Benchmarking Anticipatory Reasoning from Early Risk Cues in Video-Language Models

Sha Luo, Yogesh Prabhu, Timothy Ossowski et al. (+2 more)

With the rapid growth of video centered social media, the ability to anticipate risky events from visual data is a promising direction for ensuring public safety and preventing real world accidents. Prior work has extensively studied supervised video risk assessment across domains such as driving, protests, and natural disasters. However, many existing datasets provide models with access to the full video sequence, including the accident itself, which substantially reduces the difficulty of the task. To better reflect real world conditions, we introduce a new video understanding benchmark RiskCueBench in which videos are carefully annotated to identify a risk signal clip, defined as the earliest moment that indicates a potential safety concern. Experimental results reveal a significant gap in current systems ability to interpret evolving situations and anticipate future risky events from early visual signals, highlighting important challenges for deploying video risk prediction models in practice.
Read the full abstract

All authors Sha Luo, Yogesh Prabhu, Timothy Ossowski, Kaiping Chen, Junjie Hu

Computation and Language

A path to natural language through tokenisation and transformers

David S. Berman and Alexander G. Stapleton

Natural languages exhibit striking regularities in their statistical structure, including notably the emergence of Zipf's and Heaps' laws. Despite this, it remains broadly unclear how these properties relate to the modern tokenisation schemes used in contemporary transformer models. In this note, we analyse the information content (as measured by the Shannon entropy) of various corpora under the assumption of a Zipfian frequency distribution, and derive a closed-form expression for the slot entropy expectation value. We then empirically investigate how byte–pair encoding (BPE) transforms corpus statistics, showing that recursive applications of BPE drive token frequencies toward a Zipfian power law while inducing a characteristic growth pattern in empirical entropy. Utilizing the ability of transformers to learn context dependent token probability distributions, we train language models on corpora tokenised at varying BPE depths, revealing that the model predictive entropies increasingly agree with Zipf-derived predictions as the BPE depth increases. Attention-based diagnostics further indicate that deeper tokenisation reduces local token dependencies, bringing the empirical distribution closer to the weakly dependent (near IID) regime. Together, these results clarify how BPE acts not only as a compression mechanism but also as a statistical transform that reconstructs key informational properties of natural language.
Read the full abstract

Computation and Language

Bare-Metal Tensor Virtualization: Overcoming the Memory Wall in Edge-AI Inference on ARM64

Bugra Kilictas and Faruk Alpay

The deployment of Large Language Models (LLMs) on edge devices is fundamentally constrained by the "Memory Wall" the bottleneck where data movement latency outstrips arithmetic throughput. Standard inference runtimes often incur significant overhead through high-level abstractions, dynamic dispatch, and unaligned memory access patterns. In this work, we present a novel "Virtual Tensor Core" architecture implemented in software, optimized specifically for ARM64 microarchitectures (Apple Silicon). By bypassing standard library containers in favor of direct memory mapping (mmap) and implementing hand-tuned NEON SIMD kernels, we achieve a form of "Software-Defined Direct Memory Access (DMA)." Our proposed Tensor Virtualization Layout (TVL) guarantees 100% cache line utilization for weight matrices, while our zero-copy loader eliminates initialization latency. Experimental results on a 110M parameter model demonstrate a stable throughput of >60 tokens/second on M2 hardware. While proprietary hardware accelerators (e.g., Apple AMX) can achieve higher peak throughput, our architecture provides a fully open, portable, and deterministic reference implementation for studying the memory bottleneck on general-purpose ARM silicon, meeting the 200ms psycholinguistic latency threshold without opaque dependencies.
Read the full abstract

Computation and Language

Automated Semantic Rules Detection (ASRD) for Emergent Communication Interpretation

Bastien Vanderplaetse, Xavier Siebert and Stéphane Dupont

The field of emergent communication within multi-agent systems examines how autonomous agents can independently develop communication strategies, without explicit programming, and adapt them to varied environments. However, few studies have focused on the interpretability of emergent languages. The research exposed in this paper proposes an Automated Semantic Rules Detection (ASRD) algorithm, which extracts relevant patterns in messages exchanged by agents trained with two different datasets on the Lewis Game, which is often studied in the context of emergent communication. ASRD helps at the interpretation of the emergent communication by relating the extracted patterns to specific attributes of the input data, thereby considerably simplifying subsequent analysis.
Read the full abstract

Computation and Language

STReasoner: Empowering LLMs for Spatio-Temporal Reasoning in Time Series via Spatial-Aware Reinforcement Learning

Juntong Ni, Shiyu Wang, Qi He et al. (+2 more)

Spatio-temporal reasoning in time series involves the explicit synthesis of temporal dynamics, spatial dependencies, and textual context. This capability is vital for high-stakes decision-making in systems such as traffic networks, power grids, and disease propagation. However, the field remains underdeveloped because most existing works prioritize predictive accuracy over reasoning. To address the gap, we introduce ST-Bench, a benchmark consisting of four core tasks, including etiological reasoning, entity identification, correlation reasoning, and in-context forecasting, developed via a network SDE-based multi-agent data synthesis pipeline. We then propose STReasoner, which empowers LLM to integrate time series, graph structure, and text for explicit reasoning. To promote spatially grounded logic, we introduce S-GRPO, a reinforcement learning algorithm that rewards performance gains specifically attributable to spatial information. Experiments show that STReasoner achieves average accuracy gains between 17% and 135% at only 0.004X the cost of proprietary models and generalizes robustly to real-world data.
Read the full abstract

All authors Juntong Ni, Shiyu Wang, Qi He, Ming Jin, Wei Jin

Computation and Language

Multi-RADS Synthetic Radiology Report Dataset and Head-to-Head Benchmarking of 41 Open-Weight and Proprietary Language Models

Kartik Bose, Abhinandan Kumar, Raghuraman Soundararajan et al. (+22 more)

Background: Reporting and Data Systems (RADS) standardize radiology risk communication but automated RADS assignment from narrative reports is challenging because of guideline complexity, output-format constraints, and limited benchmarking across RADS frameworks and model sizes. Purpose: To create RXL-RADSet, a radiologist-verified synthetic multi-RADS benchmark, and compare validity and accuracy of open-weight small language models (SLMs) with a proprietary model for RADS assignment. Materials and Methods: RXL-RADSet contains 1,600 synthetic radiology reports across 10 RADS (BI-RADS, CAD-RADS, GB-RADS, LI-RADS, Lung-RADS, NI-RADS, O-RADS, PI-RADS, TI-RADS, VI-RADS) and multiple modalities. Reports were generated by LLMs using scenario plans and simulated radiologist styles and underwent two-stage radiologist verification. We evaluated 41 quantized SLMs (12 families, 0.135-32B parameters) and GPT-5.2 under a fixed guided prompt. Primary endpoints were validity and accuracy; a secondary analysis compared guided versus zero-shot prompting. Results: Under guided prompting GPT-5.2 achieved 99.8% validity and 81.1% accuracy (1,600 predictions). Pooled SLMs (65,600 predictions) achieved 96.8% validity and 61.1% accuracy; top SLMs in the 20-32B range reached ~99% validity and mid-to-high 70% accuracy. Performance scaled with model size (inflection between <1B and >=10B) and declined with RADS complexity primarily due to classification difficulty rather than invalid outputs. Guided prompting improved validity (99.2% vs 96.7%) and accuracy (78.5% vs 69.6%) compared with zero-shot. Conclusion: RXL-RADSet provides a radiologist-verified multi-RADS benchmark; large SLMs (20-32B) can approach proprietary-model performance under guided prompting, but gaps remain for higher-complexity schemes.
Read the full abstract

All authors Kartik Bose, Abhinandan Kumar, Raghuraman Soundararajan, Priya Mudgil, Samonee Ralmilay, Niharika Dutta, Manphool Singhal, Arun Kumar, Saugata Sen, Anurima Patra, Priya Ghosh, Abanti Das, Amit Gupta, Ashish Verma, Dipin Sudhakaran, Ekta Dhamija, Himangi Unde, Ishan Kumar, Krithika Rangarajan, Prerna Garg, Rachel Sequeira, Sudhin Shylendran, Taruna Yadav, Tej Pal, Pankaj Gupta

Computation and Language

MalruleLib: Large-Scale Executable Misconception Reasoning with Step Traces for Modeling Student Thinking in Mathematics

Xinghe Chen, Naiming Liu and Shashank Sonkar

Student mistakes in mathematics are often systematic: a learner applies a coherent but wrong procedure and repeats it across contexts. We introduce MalruleLib, a learning-science-grounded framework that translates documented misconceptions into executable procedures, drawing on 67 learning-science and mathematics education sources, and generates step-by-step traces of malrule-consistent student work. We formalize a core student-modeling problem as Malrule Reasoning Accuracy (MRA): infer a misconception from one worked mistake and predict the student's next answer under cross-template rephrasing. Across nine language models (4B-120B), accuracy drops from 66% on direct problem solving to 40% on cross-template misconception prediction. MalruleLib encodes 101 malrules over 498 parameterized problem templates and produces paired dual-path traces for both correct reasoning and malrule-consistent student reasoning. Because malrules are executable and templates are parameterizable, MalruleLib can generate over one million instances, enabling scalable supervision and controlled evaluation. Using MalruleLib, we observe cross-template degradations of 10-21%, while providing student step traces improves prediction by 3-15%. We release MalruleLib as infrastructure for educational AI that models student procedures across contexts, enabling diagnosis and feedback that targets the underlying misconception.
Read the full abstract

Computation and Language · Cross-listed from cs.IR

Fine-tuning Small Language Models as Efficient Enterprise Search Relevance Labelers

Yue Kang, Zhuoyi Huang, Benji Schussheim et al. (+19 more)

In enterprise search, building high-quality datasets at scale remains a central challenge due to the difficulty of acquiring labeled data. To resolve this challenge, we propose an efficient approach to fine-tune small language models (SLMs) for accurate relevance labeling, enabling high-throughput, domain-specific labeling comparable or even better in quality to that of state-of-the-art large language models (LLMs). To overcome the lack of high-quality and accessible datasets in the enterprise domain, our method leverages on synthetic data generation. Specifically, we employ an LLM to synthesize realistic enterprise queries from a seed document, apply BM25 to retrieve hard negatives, and use a teacher LLM to assign relevance scores. The resulting dataset is then distilled into an SLM, producing a compact relevance labeler. We evaluate our approach on a high-quality benchmark consisting of 923 enterprise query-document pairs annotated by trained human annotators, and show that the distilled SLM achieves agreement with human judgments on par with or better than the teacher LLM. Furthermore, our fine-tuned labeler substantially improves throughput, achieving 17 times increase while also being 19 times more cost-effective. This approach enables scalable and cost-effective relevance labeling for enterprise-scale retrieval applications, supporting rapid offline evaluation and iteration in real-world settings.
Read the full abstract

All authors Yue Kang, Zhuoyi Huang, Benji Schussheim, Diana Licon, Dina Atia, Shixing Cao, Jacob Danovitch, Kunho Kim, Billy Norcilien, Jonah Karpman, Mahmound Sayed, Mike Taylor, Tao Sun, Pavel Metrikov, Vipul Agarwal, Chris Quirk, Ye-Yi Wang, Nick Craswell, Irene Shaffer, Tianwei Chen, Sulaiman Vesal, Soundar Srinivasan

Computation and Language

UltraLogic: Enhancing LLM Reasoning through Large-Scale Data Synthesis and Bipolar Float Reward

Yile Liu, Yixian Liu, Zongwei Li et al. (+7 more)

While Large Language Models (LLMs) have demonstrated significant potential in natural language processing , complex general-purpose reasoning requiring multi-step logic, planning, and verification remains a critical bottleneck. Although Reinforcement Learning with Verifiable Rewards (RLVR) has succeeded in specific domains , the field lacks large-scale, high-quality, and difficulty-calibrated data for general reasoning. To address this, we propose UltraLogic, a framework that decouples the logical core of a problem from its natural language expression through a Code-based Solving methodology to automate high-quality data production. The framework comprises hundreds of unique task types and an automated calibration pipeline across ten difficulty levels. Furthermore, to mitigate binary reward sparsity and the Non-negative Reward Trap, we introduce the Bipolar Float Reward (BFR) mechanism, utilizing graded penalties to effectively distinguish perfect responses from those with logical flaws. Our experiments demonstrate that task diversity is the primary driver for reasoning enhancement , and that BFR, combined with a difficulty matching strategy, significantly improves training efficiency, guiding models toward global logical optima.
Read the full abstract

All authors Yile Liu, Yixian Liu, Zongwei Li, Yufei Huang, Xinhua Feng, Zhichao Hu, Jinglu Hu, Jianfeng Yan, Fengzong Lian, Yuhong Liu

Computation and Language

DIP: Dynamic In-Context Planner For Diffusion Language Models

Yang Li, Han Meng, Chenan Wang et al. (+3 more)

Diffusion language models (DLMs) have shown strong potential for general natural language tasks with in-context examples. Existing In-Context Learning (ICL) approaches largely inherit the practice of autoregressive language models (ARLMs), incorporating all examples into a fixed prompt. However, applying this rigid, static-prompt paradigm to DLMs incurs substantial computational overhead, as the model must evaluate the maximum context length at every step. We address this inefficiency with a key discovery: the block-wise KV-cache mechanism inherent to DLM inference enables the low-cost dynamic adjustment of the context. Following this intuition, our core idea is to start generation with a minimal prompt and progressively insert additional examples on the fly only when the generated tokens are of low confidence. Through rigorous empirical evaluations, we observe that average verified token confidence correlates strongly with generation accuracy, making it a reliable and computationally efficient signal of token quality. Formally, we propose Dynamic In-Context Planner (DIP), a context-optimization algorithm based on average verified confidence that dynamically ranks and inserts in-context examples during generation, rather than providing all examples up front. Experimental results on math and coding benchmarks with LLaDA-1.5 and LLaDA-8B-Instruct show that DIP achieves up to $1.59\times$ and $1.36\times$ speedups, respectively, while largely preserving the generation quality of the fixed-prompt baseline. Code: https://github.com/wmd3i/DIP
Read the full abstract

All authors Yang Li, Han Meng, Chenan Wang, Zhenyu Bi, Xuan Wang, Haipeng Chen

Computation and Language

X-MuTeST: A Multilingual Benchmark for Explainable Hate Speech Detection and A Novel LLM-consulted Explanation Framework

Mohammad Zia Ur Rehman, Sai Kartheek Reddy Kasu, Shashivardhan Reddy Koppula et al. (+3 more)

Hate speech detection on social media faces challenges in both accuracy and explainability, especially for underexplored Indic languages. We propose a novel explainability-guided training framework, X-MuTeST (eXplainable Multilingual haTe Speech deTection), for hate speech detection that combines high-level semantic reasoning from large language models (LLMs) with traditional attention-enhancing techniques. We extend this research to Hindi and Telugu alongside English by providing benchmark human-annotated rationales for each word to justify the assigned class label. The X-MuTeST explainability method computes the difference between the prediction probabilities of the original text and those of unigrams, bigrams, and trigrams. Final explanations are computed as the union between LLM explanations and X-MuTeST explanations. We show that leveraging human rationales during training enhances both classification performance and explainability. Moreover, combining human rationales with our explainability method to refine the model attention yields further improvements. We evaluate explainability using Plausibility metrics such as Token-F1 and IOU-F1 and Faithfulness metrics such as Comprehensiveness and Sufficiency. By focusing on under-resourced languages, our work advances hate speech detection across diverse linguistic contexts. Our dataset includes token-level rationale annotations for 6,004 Hindi, 4,492 Telugu, and 6,334 English samples. Data and code are available on https://github.com/ziarehman30/X-MuTeST
Read the full abstract

All authors Mohammad Zia Ur Rehman, Sai Kartheek Reddy Kasu, Shashivardhan Reddy Koppula, Sai Rithwik Reddy Chirra, Shwetank Shekhar Singh, Nagendra Kumar

Computation and Language

MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory

Shengtao Zhang, Jiaqian Wang, Ruiwen Zhou et al. (+11 more)

The hallmark of human intelligence is the self-evolving ability to master new skills by learning from past experiences. However, current AI agents struggle to emulate this self-evolution: fine-tuning is computationally expensive and prone to catastrophic forgetting, while existing memory-based methods rely on passive semantic matching that often retrieves noise. To address these challenges, we propose MemRL, a non-parametric approach that evolves via reinforcement learning on episodic memory. By decoupling stable reasoning from plastic memory, MemRL employs a Two-Phase Retrieval mechanism to filter noise and identify high-utility strategies through environmental feedback. Extensive experiments on HLE, BigCodeBench, ALFWorld, and Lifelong Agent Bench demonstrate that MemRL significantly outperforms state-of-the-art baselines, confirming that MemRL effectively reconciles the stability-plasticity dilemma, enabling continuous runtime improvement without weight updates. Code is available at https://github.com/MemTensor/MemRL.
Read the full abstract

All authors Shengtao Zhang, Jiaqian Wang, Ruiwen Zhou, Junwei Liao, Yuchen Feng, Zhuo Li, Yujie Zheng, Weinan Zhang, Ying Wen, Zhiyu Li, Feiyu Xiong, Yutao Qi, Bo Tang, Muning Wen

Computation and Language

Maximizing Local Entropy Where It Matters: Prefix-Aware Localized LLM Unlearning

Naixin Zhai, Pengyang Shao, Binbin Zheng et al. (+4 more)

Machine unlearning aims to forget sensitive knowledge from Large Language Models (LLMs) while maintaining general utility. However, existing approaches typically treat all tokens in a response indiscriminately and enforce uncertainty over the entire vocabulary. This global treatment results in unnecessary utility degradation and extends optimization to content-agnostic regions. To address these limitations, we propose PALU (Prefix-Aware Localized Unlearning), a framework driven by a local entropy maximization objective across both temporal and vocabulary dimensions. PALU reveals that (i) suppressing the sensitive prefix alone is sufficient to sever the causal generation link, and (ii) flattening only the top-$k$ logits is adequate to maximize uncertainty in the critical subspace. These findings allow PALU to alleviate redundant optimization across the full vocabulary and parameter space while minimizing collateral damage to general model performance. Comprehensive evaluations validate that PALU achieves superior forgetting efficacy and utility preservation compared to state-of-the-art baselines. Our code is available at https://github.com/nxZhai/PALU.
Read the full abstract

All authors Naixin Zhai, Pengyang Shao, Binbin Zheng, Yonghui Yang, Fei Shen, Long Bai, Xun Yang