Peer Revenue Equity in science

Papers

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

Computer Science Computation and Language

cs.CL Watch Computation and Language

Monday 16 March

Computation and Language

Extracting Breast Cancer Phenotypes from Clinical Notes: Comparing LLMs with Classical Ontology Methods

Abdullah Bin Faiz, Arbaz Khan Shehzad, Asad Afzal et al. (+5 more)

A significant amount of data held in Oncology Electronic Medical Records (EMRs) is contained in unstructured provider notes – including but not limited to the chemotherapy (or cancer treatment) outcome, different biomarkers, the tumor's location, sizes, and growth patterns of a patient. The clinical studies show that the majority of oncologists are comfortable providing these valuable insights in their notes in a natural language rather than the relevant structured fields of an EMR. The major contribution of this research is to report an LLM-based framework to process provider notes and extract valuable medical knowledge and phenotype mentioned above, with a focus on the domain of oncology. In this paper, we focus on extracting phenotypes related to breast cancer using our LLM framework, and then compare its performance with earlier works that used knowledge-driven annotation system, paired with the NCIt Ontology Annotator. The results of the study show that an LLM-based information extraction framework can be easily adapted to extract phenotypes with an accuracy that is comparable to the classical ontology-based methods. However, once trained, they could be easily fine-tuned to cater for other cancer types and diseases.
Read the full abstract

All authors Abdullah Bin Faiz, Arbaz Khan Shehzad, Asad Afzal, Momin Tariq, Muhammad Siddiqi, Muhammad Usamah Shahid, Maryam Noor Awan, Muddassar Farooq

Computation and Language

A Comparative Study of Demonstration Selection for Practical Large Language Models-based Next POI Prediction

Ryo Nishida, Masayuki Kawarada, Tatsuya Ishigaki et al. (+2 more)

This paper investigates demonstration selection strategies for predicting a user's next point-of-interest (POI) using large language models (LLMs), aiming to accurately forecast a user's subsequent location based on historical check-in data. While in-context learning (ICL) with LLMs has recently gained attention as a promising alternative to traditional supervised approaches, the effectiveness of ICL significantly depends on the selected demonstration. Although previous studies have examined methods such as random selection, embedding-based selection, and task-specific selection, there remains a lack of comprehensive comparative analysis among these strategies. To bridge this gap and clarify the best practices for real-world applications, we comprehensively evaluate existing demonstration selection methods alongside simpler heuristic approaches such as geographical proximity, temporal ordering, and sequential patterns. Extensive experiments conducted on three real-world datasets indicate that these heuristic methods consistently outperform more complex and computationally demanding embedding-based methods, both in terms of computational cost and prediction accuracy. Notably, in certain scenarios, LLMs using demonstrations selected by these simpler heuristic methods even outperform existing fine-tuned models, without requiring further training. Our source code is available at: https://github.com/ryonsd/DS-LLM4POI.
Read the full abstract

All authors Ryo Nishida, Masayuki Kawarada, Tatsuya Ishigaki, Hiroya Takamura, Masaki Onishi

Computation and Language · Cross-listed from cs.DL

Exploring Novelty Differences between Industry and Academia: A Knowledge Entity-centric Perspective

Hongye Zhao, Yi Zhao and Chengzhi Zhang

Academia and industry each possess distinct advantages in advancing technological progress. Academia's core mission is to promote open dissemination of research results and drive disciplinary progress. The industry values knowledge appropriability and core competitiveness, yet actively engages in open practices like academic conferences and platform sharing, creating a knowledge strategy paradox. Highly novel and publicly accessible knowledge serves as the driving force behind technological advancement. However, it remains unclear whether industry or academia can produce more novel research outcomes. Some studies argue that academia tends to generate more novel ideas, while others suggest that industry researchers are more likely to drive breakthroughs. Previous studies have been limited by data sources and inconsistent measures of novelty. To address these gaps, this study conducts an analysis using four types of fine-grained knowledge entities (Method, Tool, Dataset, Metric), calculates semantic distances between entities within a unified semantic space to quantify novelty, and achieves comparability of novelty across different types of literature. Then, a regression model is constructed to analyze the differences in publication novelty between industry and academia. The results indicate that academia demonstrates higher novelty outputs, which is particularly evident in patents. At the entity level, both academia and industry emphasize method-driven advancements in papers, while industry holds a unique advantage in datasets. Additionally, academia-industry collaboration has a limited effect on enhancing the novelty of research papers, but it helps to enhance the novelty of patents. We release our data and associated codes at https://github.com/tinierZhao/entity_novelty.
Read the full abstract

Computation and Language

Understanding Moral Reasoning Trajectories in Large Language Models: Toward Probing-Based Explainability

Fan Huang, Haewoon Kwak and Jisun An

Large language models (LLMs) increasingly participate in morally sensitive decision-making, yet how they organize ethical frameworks across reasoning steps remains underexplored. We introduce moral reasoning trajectories, sequences of ethical framework invocations across intermediate reasoning steps, and analyze their dynamics across six models and three benchmarks. We find that moral reasoning involves systematic multi-framework deliberation: 55.4–57.7% of consecutive steps involve framework switches, and only 16.4–17.8% of trajectories remain framework-consistent. Unstable trajectories remain 1.29 times more susceptible to persuasive attacks (p=0.015). At the representation level, linear probes localize framework-specific encoding to model-specific layers (layer 63/81 for Llama-3.3-70B; layer 17/81 for Qwen2.5-72B), achieving 16.8–22.2% lower KL divergence than the step-prior baseline. Activation steering applied during generation moves the framework-consistency–accuracy relationship, widening it for Qwen2.5-72B and erasing it for Llama-3.3-70B, and a probe-space layer sweep bounds the attainable drift reduction at 6.7–8.9%. We further propose a Moral Representation Consistency (MRC) metric whose underlying framework attributions are validated by human annotators (mean cosine similarity = 0.859), and we report what an automated coherence rater does and does not establish about it.
Read the full abstract

Computation and Language · Cross-listed from cs.SE

FormulaCode: Evaluating Agentic Optimization on Large Codebases

Atharva Sehgal, James Hou, Akanksha Sarkar et al. (+4 more)

Large language model (LLM) coding agents increasingly operate at the repository level, motivating benchmarks that evaluate their ability to optimize entire codebases under realistic constraints. Existing code benchmarks largely rely on synthetic tasks, binary correctness signals, or single-objective evaluation, limiting their ability to assess holistic optimization behavior. We introduce FormulaCode, a benchmark for evaluating agentic optimization on large, real-world codebases with fine-grained, multi-objective performance metrics. FormulaCode comprises 957 performance bottlenecks mined from scientific Python repositories on GitHub, each paired with expert-authored patches and, on average, 264.6 community-maintained performance workloads per task, enabling the holistic ability of LLM agents to optimize codebases under realistic correctness and performance constraints. Our evaluations reveal that repository-scale, multi-objective optimization remains a major challenge for frontier LLM agents. Project website at: https://formula-code.github.io
Read the full abstract

All authors Atharva Sehgal, James Hou, Akanksha Sarkar, Ishaan Mantripragada, Swarat Chaudhuri, Jennifer J. Sun, Yisong Yue

Computation and Language

RadAnnotate: Large Language Models for Efficient and Reliable Radiology Report Annotation

Saisha Pradeep Shetty, Roger Eric Goldman and Vladimir Filkov

Radiology report annotation is essential for clinical NLP, yet manual labeling is slow and costly. We present RadAnnotate, an LLM-based framework that studies retrieval-augmented synthetic reports and confidence-based selective automation to reduce expert effort for labeling in RadGraph. We study RadGraph-style entity labeling (graph nodes) and leave relation extraction (edges) to future work. First, we train entity-specific classifiers on gold-standard reports and characterize their strengths and failure modes across anatomy and observation categories, with uncertain observations hardest to learn. Second, we generate RAG-guided synthetic reports and show that synthetic-only models remain within 1-2 F1 points of gold-trained models, and that synthetic augmentation is especially helpful for uncertain observations in a low-resource setting, improving F1 from 0.61 to 0.70. Finally, by learning entity-specific confidence thresholds, RadAnnotate can automatically annotate 55-90% of reports at 0.86-0.92 entity match score while routing low-confidence cases for expert review.
Read the full abstract

Computation and Language · Cross-listed from cs.CV

Mostly Text, Smart Visuals: Asymmetric Text-Visual Pruning for Large Vision-Language Models

Sijie Li, Biao Qian and Jungong Han

Network pruning is an effective technique for enabling lightweight Large Vision-Language Models (LVLMs), which primarily incorporates both weights and activations into the importance metric. However, existing efforts typically process calibration data from different modalities in a unified manner, overlooking modality-specific behaviors. This raises a critical challenge: how to address the divergent behaviors of textual and visual tokens for accurate pruning of LVLMs. To this end, we systematically investigate the sensitivity of visual and textual tokens to the pruning operation by decoupling their corresponding weights, revealing that: (i) the textual pathway should be calibrated via text tokens, since it exhibits higher sensitivity than the visual pathway; (ii) the visual pathway exhibits high redundancy, permitting even 50% sparsity. Motivated by these insights, we propose a simple yet effective Asymmetric Text-Visual Weight Pruning method for LVLMs, dubbed ATV-Pruning, which establishes the importance metric for accurate weight pruning by selecting the informative tokens from both textual and visual pathways. Specifically, ATV-Pruning integrates two primary innovations: first, a calibration pool is adaptively constructed by drawing on all textual tokens and a subset of visual tokens; second, we devise a layer-adaptive selection strategy to yield important visual tokens. Finally, extensive experiments across standard multimodal benchmarks verify the superiority of our ATV-Pruning over state-of-the-art methods.
Read the full abstract

Computation and Language

NLP Occupational Emergence Analysis: How Occupations Form and Evolve in Real Time – A Zero-Assumption Method Demonstrated on AI in the US Technology Workforce, 2022-2026

David Nordfors

Occupations form and evolve faster than classification systems can track. We propose that a genuine occupation is a self-reinforcing structure (a bipartite co-attractor) in which a shared professional vocabulary makes practitioners cohesive as a group, and the cohesive group sustains the vocabulary. This co-attractor concept enables a zero-assumption method for detecting occupational emergence from resume data, requiring no predefined taxonomy or job titles: we test vocabulary cohesion and population cohesion independently, with ablation to test whether the vocabulary is the mechanism binding the population. Applied to 8.2 million US resumes (2022-2026), the method correctly identifies established occupations and reveals a striking asymmetry for AI: a cohesive professional vocabulary formed rapidly in early 2024, but the practitioner population never cohered. The pre-existing AI community dissolved as the tools went mainstream, and the new vocabulary was absorbed into existing careers rather than binding a new occupation. AI appears to be a diffusing technology, not an emerging occupation. We discuss whether introducing an "AI Engineer" occupational category could catalyze population cohesion around the already-formed vocabulary, completing the co-attractor.
Read the full abstract

Computation and Language · Cross-listed from cs.MM

Visual Set Program Synthesizer

Zehua Cheng, Wei Dai, Wenhu Zhang et al. (+2 more)

A user pointing their phone at a supermarket shelf and asking "Which soda has the least sugar?" poses a difficult challenge for current visual Al assistants. Such queries require not only object recognition, but explicit set-based reasoning such as filtering, comparison, and aggregation. Standard endto-end MLLMs often fail at these tasks because they lack an explicit mechanism for compositional logic. We propose treating visual reasoning as Visual Program Synthesis, where the model first generates a symbolic program that is executed by a separate engine grounded in visual scenes. We also introduce Set-VQA, a new benchmark designed specifically for evaluating set-based visual reasoning. Experiments show that our approach significantly outperforms state-of-the-art baselines on complex reasoning tasks, producing more systematic and transparent behavior while substantially improving answer accuracy. These results demonstrate that program-driven reasoning provides a principled alternative to black-box visual-language inference.
Read the full abstract

All authors Zehua Cheng, Wei Dai, Wenhu Zhang, Thomas Lukasiewicz, Jiahao Sun

Computation and Language

Aligning Paralinguistic Understanding and Generation in Speech LLMs via Multi-Task Reinforcement Learning

Jingxiang Chen, Minseok Kim, Seong-Gyun Leem et al. (+13 more)

Speech large language models (LLMs) observe paralinguistic cues such as prosody, emotion, and non-verbal sounds–crucial for intent understanding. However, leveraging these cues faces challenges: limited training data, annotation difficulty, and models exploiting lexical shortcuts over paralinguistic signals. We propose multi-task reinforcement learning (RL) with chain-of-thought prompting that elicits explicit affective reasoning. To address data scarcity, we introduce a paralinguistics-aware speech LLM (PALLM) that jointly optimizes sentiment classification from audio and paralinguistics-aware response generation via a two-stage pipeline. Experiments demonstrate that our approach improves paralinguistics understanding over both supervised baselines and strong proprietary models (Gemini-2.5-Pro, GPT-4o-audio) by 8-12% on Expresso, IEMOCAP, and RAVDESS. The results show that modeling paralinguistic reasoning with multi-task RL is crucial for building emotionally intelligent speech LLMs.
Read the full abstract

All authors Jingxiang Chen, Minseok Kim, Seong-Gyun Leem, Yin Huang, Rashi Rungta, Zhicheng Ouyang, Haibin Wu, Surya Teja Appini, Ankur Bansal, Yang Bai, Yue Liu, Florian Metze, Ahmed A Aly, Anuj Kumar, Ariya Rastrow, Zhaojiang Lin

Computation and Language

Robust Language Identification for Romansh Varieties

Charlotte Model, Sina Ahmadi and Jannis Vamvas

The Romansh language has several regional varieties, called idioms, which sometimes have limited mutual intelligibility. Despite this linguistic diversity, there has been a lack of documented efforts to build a language identification (LID) system that can distinguish between these idioms. Since Romansh LID should also be able to recognize Rumantsch Grischun, a supra-regional variety that combines elements of several idioms, this makes for a novel and interesting classification problem. In this paper, we present a LID system for Romansh idioms based on an SVM approach. We evaluate our model on a newly curated benchmark across two domains and find that it reaches an average in-domain accuracy of 97%, enabling applications such as idiom-aware spell checking or machine translation. Our classifier is publicly available.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

MAC: Multi-Agent Constitution Learning

Rushil Thareja, Gautam Gupta, Francesco Pinto et al. (+1 more)

Constitutional AI is a method to oversee and control LLMs based on a set of rules written in natural language. These rules are typically written by human experts, but could in principle be learned automatically given sufficient training data for the desired behavior. Existing LLM-based prompt optimizers attempt this but are ineffective at learning constitutions since (i) they require many labeled examples and (ii) lack structure in the optimized prompts, leading to diminishing improvements as prompt size grows. To address these limitations, we propose Multi-Agent Constitutional Learning (MAC), which optimizes over structured prompts represented as sets of rules using a network of agents with specialized tasks to accept, edit, or reject rule updates. We also present MAC+, which improves performance by training agents on successful trajectories to reinforce updates leading to higher reward. We evaluate MAC on tagging Personally Identifiable Information (PII), a classification task with limited labels where interpretability is critical, and demonstrate that it generalizes to other agentic tasks such as tool calling. MAC outperforms recent prompt optimization methods by over 50%, produces human-readable and auditable rule sets, and achieves performance comparable to supervised fine-tuning and GRPO without requiring parameter updates.
Read the full abstract

All authors Rushil Thareja, Gautam Gupta, Francesco Pinto, Nils Lukas

Computation and Language

MoLoRA: Composable Specialization via Per-Token Adapter Routing

Shrey Shah and Justin Wagle

Multi-adapter serving systems route entire sequences to a single adapter, forcing a choice when requests span multiple domains. This assumption fails in two important settings: (1) multimodal generation, where text and image tokens require different adapters within the same sequence, and (2) mixed-capability requests like "write code to solve this equation," which need expertise from multiple specialized adapters. We introduce per-token routing, which routes individual tokens to adapters based on either vocabulary structure (for multimodal models) or learned gating (for semantic specialization). Per-token routing is provably optimal, achieving work N for N tokens versus K \cdot N for per-sequence routing with K adapter types. Our key contribution is MoLoRA (Mixture of LoRA), which enables composable specialization: load multiple domain-specific adapters and let a learned router select the appropriate adapter per-token. We demonstrate that specialization dramatically beats scale: MoLoRA enables Qwen3-1.7B to exceed Qwen3-8B across four reasoning benchmarks while being 4.7x smaller. This enables modular expertise at inference time: train focused LoRAs independently, combine them without retraining, and add new capabilities by simply loading new adapters.
Read the full abstract

Computation and Language

A Family of LLMs Liberated from Static Vocabularies

Aleph Alpha: Adnen Abdessaied, Artur Baranowski, Lukas Balles et al. (+33 more)

Tokenization is a central component of natural language processing in current large language models (LLMs), enabling models to convert raw text into processable units. Although learned tokenizers are widely adopted, they exhibit notable limitations, including their large, fixed vocabulary sizes and poor adaptability to new domains or languages. We present a family of models with up to 70 billion parameters based on the hierarchical autoregressive transformer (HAT) architecture. In HAT, an encoder transformer aggregates bytes into word embeddings and then feeds them to the backbone, a classical autoregressive transformer. The outputs of the backbone are then cross-attended by the decoder and converted back into bytes. We show that we can reuse available pre-trained models by converting the Llama 3.1 8B and 70B models into the HAT architecture: Llama-3.1-8B-TFree-HAT and Llama-3.1-70B-TFree-HAT are byte-level models whose encoder and decoder are trained from scratch, but where we adapt the pre-trained Llama backbone, i.e., the transformer blocks with the embedding matrix and head removed, to handle word embeddings instead of the original tokens. We also provide a 7B HAT model, Llama-TFree-HAT-Pretrained, trained entirely from scratch on nearly 4 trillion words. The HAT architecture improves text compression by reducing the number of required sequence positions and enhances robustness to intra-word variations, e.g., spelling differences. Through pre-training, as well as subsequent supervised fine-tuning and direct preference optimization in English and German, we show strong proficiency in both languages, improving on the original Llama 3.1 in most benchmarks. We release our models (including 200 pre-training checkpoints) on Hugging Face.
Read the full abstract

All authors Aleph Alpha: Adnen Abdessaied, Artur Baranowski, Lukas Balles, Michael Barlow, Fabien C. Y. Benureau, Felix Berkenkamp, Lukas Bluebaum, Bastian Boll, Thomas F. Burns, Björn Deiseroth, Constantin Eichenberg, David Friede, Pablo Iyu Guerrero, Ahmed Hammam, Bastian Harren, Johann Higl, Yasser Jadidi, Carina Kauf, Johannes Messner, Jan Hendrik Metzen, Max Meuer, Vedant Nanda, Pit Neitemeier, Koen Oostermeijer, Letitia Parcalabescu, Markus Pernpointner, Felix Reinfurt, Dylan Rodriquez, Grégory Schott, Philipp Siedler, Martin Simonovsky, Till Speicher, Volker Stampa, Stephan Wäldchen, Samuel Weinbach, Gregor Ziegltrum

Computation and Language

POLAR:A Per-User Association Test in Embedding Space

Pedro Bento, Arthur Buzelin, Arthur Chagas et al. (+7 more)

Most intrinsic association probes operate at the word, sentence, or corpus level, obscuring author-level variation. We present POLAR (Per-user On-axis Lexical Association Re-port), a per-user lexical association test that runs in the embedding space of a lightly adapted masked language model. Authors are represented by private deterministic to-kens; POLAR projects these vectors onto curated lexicalaxes and reports standardized effects with permutation p-values and Benjamini–Hochberg control. On a balanced bot–human Twitter benchmark, POLAR cleanly separates LLM-driven bots from organic accounts; on an extremist forum,it quantifies strong alignment with slur lexicons and reveals rightward drift over time. The method is modular to new attribute sets and provides concise, per-author diagnostics for computational social science. All code is publicly avail-able at https://github.com/pedroaugtb/POLAR-A-Per-User-Association-Test-in-Embedding-Space.
Read the full abstract

All authors Pedro Bento, Arthur Buzelin, Arthur Chagas, Yan Aquino, Victoria Estanislau, Samira Malaquias, Pedro Robles Dutenhefner, Gisele L. Pappa, Virgilio Almeida, Wagner MeiraJr

Computation and Language

BanglaSocialBench: A Benchmark for Evaluating Sociopragmatic and Cultural Alignment of LLMs in Bangladeshi Social Interaction

Tanvir Ahmed Sijan, S. M Golam Rifat, Pankaj Chowdhury Partha et al. (+2 more)

Large Language Models have demonstrated strong multilingual fluency, yet fluency alone does not guarantee socially appropriate language use. In high-context languages, communicative competence requires sensitivity to social hierarchy, relational roles, and interactional norms that are encoded directly in everyday language. Bangla exemplifies this challenge through its three-tiered pronominal system, kinship-based addressing, and culturally embedded social customs. We introduce BanglaSocialBench, the first benchmark designed to evaluate sociopragmatic competence in Bangla through context-dependent language use rather than factual recall. The benchmark spans three domains: Bangla Address Terms, Kinship Reasoning, and Social Customs, comprising 1,719 culturally grounded instances written and verified by native Bangla speakers. We evaluate twelve contemporary LLMs in a zero-shot setting and observe systematic patterns of cultural misalignment. Models frequently default to overly formal address forms, fail to recognize multiple socially acceptable address pronouns, and conflate kinship terminology across religious contexts. Our findings show that sociopragmatic failures are often structured and non-random; for example, inappropriate addressing choices concentrate heavily in downward-hierarchy (Elder$\rightarrow$Younger) and informal contexts. This reveals persistent limitations in how current LLMs infer and apply culturally appropriate language use in realistic Bangladeshi social interactions.
Read the full abstract

All authors Tanvir Ahmed Sijan, S. M Golam Rifat, Pankaj Chowdhury Partha, Md. Tanjeed Islam, Md. Musfique Anwar

Computation and Language

CTG-DB: An Ontology-Based Transformation of ClinicalTrials.gov to Enable Cross-Trial Drug Safety Analyses

Jeffery L. Painter, François Haguinet and Andrew Bate

ClinicalTrials .gov (CT .gov) is the largest publicly accessible registry of clinical studies, yet its registry-oriented architecture and heterogeneous adverse event (AE) terminology limit systematic pharmacovigilance (PV) analytics. AEs are typically recorded as investigator-reported text rather than standardized identifiers, requiring manual reconciliation to identify coherent safety concepts. We present the ClinicalTrials .gov Transformation Database (CTG-DB), an open-source pipeline that ingests the complete CT .gov XML archive and produces a relational database aligned to standardized AE terminology using the Medical Dictionary for Regulatory Activities (MedDRA). CTG-DB preserves arm-level denominators, represents placebo and comparator arms, and normalizes AE terminology using deterministic exact and fuzzy matching to ensure transparent and reproducible mappings. This framework enables concept-level retrieval and cross-trial aggregation for scalable placebo-referenced safety analyses and integration of clinical trial evidence into downstream PV signal detection.
Read the full abstract

Computation and Language · Cross-listed from cs.HC

Machine Translation in the Wild: User Reaction to Xiaohongshu's Built-In Translation Feature

Sui He

This paper examines user reactions to the launch of the machine translation (MT) feature on Xiaohongshu, a Chinese social media and e-commerce platform, in January 2025. Drawing on a dataset of 6,723 comments collected from 11 official posts promoting the translation function, this paper combines sentiment analysis with thematic analysis to investigate how users perceived and experimented with the function. Results show that reactions were generally positive, although concerns regarding functionality, accessibility, and translation accuracy were also expressed. In addition, users actively tested the function with inputs that are atypical for everyday online communication, including stand-alone words and phrases, abbreviations, internet slang, and symbolic or encoded forms. Successful decoding of these texts elicited positive responses, while testing of more conventional language remained fairly limited. This could lead to uncritical acceptance of MT outputs by users, highlighting the importance of closer collaboration among computer scientists, translation scholars, and platform designers to improve MT performance and promote informed user engagement in real-world scenarios.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

Prompt Engineering for Scale Development in Generative Psychometrics

Lara Lee Russell-Lasalandra and Hudson Golino

This Monte Carlo simulation examines how prompt engineering strategies shape the quality of large language model (LLM)–generated personality assessment items within the AI-GENIE framework for generative psychometrics. Item pools targeting the Big Five traits were generated using multiple prompting designs (zero-shot, few-shot, persona-based, and adaptive), model temperatures, and LLMs, then evaluated and reduced using network psychometric methods. Across all conditions, AI-GENIE reliably improved structural validity following reduction, with the magnitude of its incremental contribution inversely related to the quality of the incoming item pool. Prompt design exerted a substantial influence on both pre- and post-reduction item quality. Adaptive prompting consistently outperformed non-adaptive strategies by sharply reducing semantic redundancy, elevating pre-reduction structural validity, and preserving substantially larger item pool, particularly when paired with newer, higher-capacity models. These gains were robust across temperature settings for most models, indicating that adaptive prompting mitigates common trade-offs between creativity and psychometric coherence. An exception was observed for the GPT-4o model at high temperatures, suggesting model-specific sensitivity to adaptive constraints at elevated stochasticity. Overall, the findings demonstrate that adaptive prompting is the strongest approach in this context, and that its benefits scale with model capability, motivating continued investigation of model–prompt interactions in generative psychometric pipelines.
Read the full abstract

Computation and Language

Agent-based imitation dynamics can yield efficiently compressed population-level vocabularies

Nathaniel Imel, Richard Futrell, Michael Franke et al. (+1 more)

Natural languages have been argued to evolve under pressure to efficiently compress meanings into words by optimizing the Information Bottleneck (IB) complexity-accuracy tradeoff. However, the underlying social dynamics that could drive the optimization of a language's vocabulary towards efficiency remain largely unknown. In parallel, evolutionary game theory has been invoked to explain the emergence of language from rudimentary agent-level dynamics, but it has not yet been tested whether such an approach can lead to efficient compression in the IB sense. Here, we provide a unified model integrating evolutionary game theory with the IB framework and show how near-optimal compression can arise in a population through an independently motivated dynamic of imprecise strategy imitation in signaling games. We find that key parameters of the model – namely, those that regulate precision in these games, as well as players' tendency to confuse similar states – lead to constrained variation of the tradeoffs achieved by emergent vocabularies. Our results suggest that evolutionary game dynamics could potentially provide a mechanistic basis for the evolution of vocabularies with information-theoretically optimal and empirically attested properties.
Read the full abstract

All authors Nathaniel Imel, Richard Futrell, Michael Franke, Noga Zaslavsky

Computation and Language

COGNAC at SemEval-2026 Task 5: LLM Ensembles for Human-Level Word Sense Plausibility Rating in Challenging Narratives

Azwad Anjum Islam and Tisa Islam Erana

We describe our system for SemEval-2026 Task 5, which requires rating the plausibility of given word senses of homonyms in short stories on a 5-point Likert scale. Systems are evaluated by the unweighted average of accuracy (within one standard deviation of mean human judgments) and Spearman Rank Correlation. We explore three prompting strategies using multiple closed-source commercial LLMs: (i) a baseline zero-shot setup, (ii) Chain-of-Thought (CoT) style prompting with structured reasoning, and (iii) a comparative prompting strategy for evaluating candidate word senses simultaneously. Furthermore, to account for the substantial inter-annotator variation present in the gold labels, we propose an ensemble setup by averaging model predictions. Our best official system, comprising an ensemble of LLMs across all three prompting strategies, placed 4th on the competition leaderboard with 0.88 accuracy and 0.83 Spearman's rho (0.86 average). Post-competition experiments with additional models further improved this performance to 0.92 accuracy and 0.85 Spearman's rho (0.89 average). We find that comparative prompting consistently improved performance across model families, and model ensembling significantly enhanced alignment with mean human judgments, suggesting that LLM ensembles are especially well suited for subjective semantic evaluation tasks involving multiple annotators.
Read the full abstract

Computation and Language · Cross-listed from cs.IR

Temporal Fact Conflicts in LLMs: Reproducibility Insights from Unifying DYNAMICQA and MULAN

Ritajit Dey, Iadh Ounis, Graham McDonald et al. (+1 more)

Large Language Models (LLMs) often struggle with temporal fact conflicts due to outdated or evolving information in their training data. Two recent studies with accompanying datasets report opposite conclusions on whether external context can effectively resolve such conflicts. DYNAMICQA evaluates how effective external context is in shifting the model's output distribution, finding that temporal facts are more resistant to change. In contrast, MULAN examines how often external context changes memorised facts, concluding that temporal facts are easier to update. In this reproducibility paper, we first reproduce experiments from both benchmarks. We then reproduce the experiments of each study on the dataset of the other to investigate the source of their disagreement. To enable direct comparison of findings, we standardise both datasets to align with the evaluation settings of each study. Importantly, using an LLM, we synthetically generate realistic natural language contexts to replace MULAN's programmatically constructed statements when reproducing the findings of DYNAMICQA. Our analysis reveals strong dataset dependence: MULAN's findings generalise under both methodological frameworks, whereas applying MULAN's evaluation to DYNAMICQA yields mixed outcomes. Finally, while the original studies only considered 7B LLMs, we reproduce these experiments across LLMs of varying sizes, revealing how model size influences the encoding and updating of temporal facts. Our results highlight how dataset design, evaluation metrics, and model size shape LLM behaviour in the presence of temporal knowledge conflicts.
Read the full abstract

All authors Ritajit Dey, Iadh Ounis, Graham McDonald, Yashar Moshfeghi

Computation and Language · Cross-listed from cs.LG

FlashSampling: Fast and Memory-Efficient Exact Sampling

Tomas Ruiz, Zhen Qin, Yifan Zhang et al. (+3 more)

Sampling from a categorical distribution is mathematically simple, but in large-vocabulary decoding, it often triggers extra memory traffic and extra kernels after the LM head. We present FlashSampling, an exact sampling primitive that fuses sampling into the LM-head matmul and never materializes the logits tensor in HBM. The method is simple: compute logits tile-by-tile on chip, add Gumbel noise, keep only one maximizer per row and per vocabulary tile, and finish with a small reduction over tiles. In tensor-parallel decoding, FlashSampling replaces the all-gather of logits with streaming peer-to-peer writes: This overlaps GPU-to-GPU communication with computation and HBM loads across up to 8 GPUs, with near-ideal scaling at large batch sizes. Our kernel is exact because $\text{argmax}$ decomposes over partitions; grouped variants for online and tensor-parallel settings are exact by hierarchical factorization of the categorical distribution. FlashSampling demonstrates kernel-level speedups on decode workloads across 4 different datacenter GPUs (H100, H200, B200, B300), and in end-to-end vLLM experiments, it reduces time per output token by up to $10%$ on the models we test. These results show that exact sampling, with no approximation, can be integrated into the matmul itself, consolidating the bandwidth-bound sampling step in an efficient epilogue.
Read the full abstract

All authors Tomas Ruiz, Zhen Qin, Yifan Zhang, Xuyang Shen, Yiran Zhong, Mengdi Wang

Computation and Language · Cross-listed from cs.LG

When Stability Fails: Hidden Failure Modes Of LLMS in Data-Constrained Scientific Decision-Making

Nazia Riasat

Large language models (LLMs) are increasingly used as decision-support tools in data-constrained scientific workflows, where correctness and validity are critical. However, evaluation practices often emphasize stability or reproducibility across repeated runs. While these properties are desirable, stability alone does not guar- antee agreement with statistical ground truth when such references are available. We introduce a controlled behavioral evaluation framework that explicitly sep- arates four dimensions of LLM decision-making: stability, correctness, prompt sensitivity, and output validity under fixed statistical inputs. We evaluate multi- ple LLMs using a statistical gene prioritization task derived from differential ex- pression analysis across prompt regimes involving strict and relaxed significance thresholds, borderline ranking scenarios, and minor wording variations. Our ex- periments show that LLMs can exhibit near-perfect run-to-run stability while sys- tematically diverging from statistical ground truth, over-selecting under relaxed thresholds, responding sharply to minor prompt wording changes, or producing syntactically plausible gene identifiers absent from the input table. Although sta- bility reflects robustness across repeated runs, it does not guarantee agreement with statistical ground truth in structured scientific decision tasks. These findings highlight the importance of explicit ground-truth validation and output validity checks when deploying LLMs in automated or semi-automated scientific work- flows.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

Persona-Conditioned Risk Behavior in Large Language Models: A Simulated Gambling Study with GPT-4.1

Sankalp Dubedy

Large language models (LLMs) are increasingly deployed as autonomous agents in uncertain, sequential decision-making contexts. Yet it remains poorly understood whether the behaviors they exhibit in such environments reflect principled cognitive patterns or simply surface-level prompt mimicry. This paper presents a controlled experiment in which GPT-4.1 was assigned one of three socioeconomic personas (Rich, Middle-income, and Poor) and placed in a structured slot-machine environment with three distinct machine configurations: Fair (50%), Biased Low (35%), and Streak (dynamic probability increasing after consecutive losses). Across 50 independent iterations per condition and 6,950 recorded decisions, we find that the model reproduces key behavioral signatures predicted by Kahneman and Tversky's Prospect Theory without being instructed to do so. The Poor persona played a mean of 37.4 rounds per session (SD=15.5) compared to 1.1 rounds for the Rich persona (SD=0.31), a difference that is highly significant (Kruskal-Wallis H=393.5, p<2.2e-16). Risk scores by persona show large effect sizes (Cohen's d=4.15 for Poor vs Rich). Emotional labels appear to function as post-hoc annotations rather than decision drivers (chi-square=3205.4, Cramer's V=0.39), and belief-updating across rounds is negligible (Spearman rho=0.032 for Poor persona, p=0.016). These findings carry implications for LLM agent design, interpretability research, and the broader question of whether classical cognitive economic biases are implicitly encoded in large-scale pretrained language models.
Read the full abstract

Computation and Language · Cross-listed from cs.CV

Evolving Contextual Safety in Multi-Modal Large Language Models via Inference-Time Self-Reflective Memory

Ce Zhang, Jinxi He, Junyi He et al. (+2 more)

Multi-modal Large Language Models (MLLMs) have achieved remarkable performance across a wide range of visual reasoning tasks, yet their vulnerability to safety risks remains a pressing concern. While prior research primarily focuses on jailbreak defenses that detect and refuse explicitly unsafe inputs, such approaches often overlook contextual safety, which requires models to distinguish subtle contextual differences between scenarios that may appear similar but diverge significantly in safety intent. In this work, we present MM-SafetyBench++, a carefully curated benchmark designed for contextual safety evaluation. Specifically, for each unsafe image-text pair, we construct a corresponding safe counterpart through minimal modifications that flip the user intent while preserving the underlying contextual meaning, enabling controlled evaluation of whether models can adapt their safety behaviors based on contextual understanding. Further, we introduce EchoSafe, a training-free framework that maintains a self-reflective memory bank to accumulate and retrieve safety insights from prior interactions. By integrating relevant past experiences into current prompts, EchoSafe enables context-aware reasoning and continual evolution of safety behavior during inference. Extensive experiments on various multi-modal safety benchmarks demonstrate that EchoSafe consistently achieves superior performance, establishing a strong baseline for advancing contextual safety in MLLMs. All benchmark data and code are available at https://echosafe-mllm.github.io.
Read the full abstract

All authors Ce Zhang, Jinxi He, Junyi He, Katia Sycara, Yaqi Xie

Computation and Language

Morphemes Without Borders: Evaluating Root-Pattern Morphology in Arabic Tokenizers and LLMs

Yara Alakeel, Chatrine Qwaider, Hanan Aldarmaki et al. (+1 more)

This work investigates how effectively large language models (LLMs) and their tokenization schemes represent and generate Arabic root-pattern morphology, probing whether they capture genuine morphological structure or rely on surface memorization. Arabic morphological system provides a rich testbed for analyzing how LLMs handle complex, non-concatenative forms and how tokenization choices influence this process. Our study begins with an evaluation of morphological fidelity across Arabic and multilingual tokenizers against gold-standard segmentation, followed by an analysis of LLM performance in productive root-pattern generation using a newly developed test set. Our findings across seven Arabic-centric and multilingual LLMs and their respective tokenizers reveal that tokenizer morphological alignment is not necessary nor sufficient for morphological generation, which questions the role of morphological tokenization in downstream performance.
Read the full abstract

All authors Yara Alakeel, Chatrine Qwaider, Hanan Aldarmaki, Sawsan Alqahtani

Computation and Language

MiroThinker-1.7 & H1: Towards Heavy-Duty Research Agents via Verification

MiroMind Team: S. Bai, L. Bing, L. Lei et al. (+40 more)

We present MiroThinker-1.7, a new research agent designed for complex long-horizon reasoning tasks. Building on this foundation, we further introduce MiroThinker-H1, which extends the agent with heavy-duty reasoning capabilities for more reliable multi-step problem solving. In particular, MiroThinker-1.7 improves the reliability of each interaction step through an agentic mid-training stage that emphasizes structured planning, contextual reasoning, and tool interaction. This enables more effective multi-step interaction and sustained reasoning across complex tasks. MiroThinker-H1 further incorporates verification directly into the reasoning process at both local and global levels. Intermediate reasoning decisions can be evaluated and refined during inference, while the overall reasoning trajectory is audited to ensure that final answers are supported by coherent chains of evidence. Across benchmarks covering open-web research, scientific reasoning, and financial analysis, MiroThinker-H1 achieves state-of-the-art performance on deep research tasks while maintaining strong results on specialized domains. We also release MiroThinker-1.7 and MiroThinker-1.7-mini as open-source models, providing competitive research-agent capabilities with significantly improved efficiency.
Read the full abstract

All authors MiroMind Team: S. Bai, L. Bing, L. Lei, R. Li, X. Li, X. Lin, E. Min, L. Su, B. Wang, L. Wang, L. Wang, S. Wang, X. Wang, Y. Zhang, Z. Zhang, G. Chen, L. Chen, Z. Cheng, Y. Deng, Z. Huang, D. Ng, J. Ni, Q. Ren, X. Tang, B.L. Wang, H. Wang, N. Wang, C. Wei, Q. Wu, J. Xia, Y. Xiao, H. Xu, X. Xu, C. Xue, Z. Yang, Z. Yang, F. Ye, H. Ye, J. Yu, C. Zhang, W. Zhang, H. Zhao, P. Zhu

Computation and Language

Mixture-of-Depths Attention

Lianghui Zhu, Yuxin Fang, Bencheng Liao et al. (+10 more)

Scaling depth is a key driver for large language models (LLMs). Yet, as LLMs become deeper, they often suffer from signal degradation: informative features formed in shallow layers are gradually diluted by repeated residual updates, making them harder to recover in deeper layers. We introduce mixture-of-depths attention (MoDA), a mechanism that allows each attention head to attend to sequence KV pairs at the current layer and depth KV pairs from preceding layers. We further describe a hardware-efficient algorithm for MoDA that resolves non-contiguous memory-access patterns, achieving 97.3% of FlashAttention-2's efficiency at a sequence length of 64K. Experiments on 1.5B-parameter models demonstrate that MoDA consistently outperforms strong baselines. Notably, it improves average perplexity by 0.2 across 10 validation benchmarks and increases average performance by 2.11% on 10 downstream tasks, with a negligible 3.7% FLOPs computational overhead. We also find that combining MoDA with post-norm yields better performance than using it with pre-norm. These results suggest that MoDA is a promising primitive for depth scaling. Code is released at https://github.com/hustvl/MoDA .
Read the full abstract

All authors Lianghui Zhu, Yuxin Fang, Bencheng Liao, Shijie Wang, Tianheng Cheng, Zilong Huang, Chen Chen, Lai Wei, Yutao Zeng, Ya Wang, Yi Lin, Yu Li, Xinggang Wang

Computation and Language

Mechanistic Origin of Moral Indifference in Language Models

Lingyu Li, Yan Teng and Yingchun Wang

Existing behavioral alignment techniques for Large Language Models (LLMs) often neglect the discrepancy between surface compliance and internal unaligned representations, leaving LLMs vulnerable to long-tail risks. More crucially, we posit that LLMs possess an inherent state of moral indifference due to compressing distinct moral concepts into uniform probability distributions. We verify and remedy this indifference in LLMs' latent representations, utilizing 251k moral vectors constructed upon Prototype Theory and the Social-Chemistry-101 dataset. Firstly, our analysis across 23 models reveals that current LLMs fail to represent the distinction between opposed moral categories and fine-grained typicality gradients within these categories; notably, neither model scaling, architecture, nor explicit alignment reshapes this indifference. We then employ Sparse Autoencoders on Qwen3-8B, isolate mono-semantic moral features, and targetedly reconstruct their topological relationships to align with ground-truth moral vectors. This representational alignment naturally improves moral reasoning and granularity, achieving a 75% pairwise win-rate on the independent adversarial Flames benchmark. Finally, we elaborate on the remedial nature of current intervention methods from an experientialist philosophy, arguing that endogenously aligned AI might require a transformation from post-hoc corrections to proactive cultivation.
Read the full abstract