arXiv_CL
-
Looped Uniform Diffusion Language Models
2026-10-21 17:24:35Saiteja Utpala, Chieh-Hsin Lai, Yuki Mitsufuji, Praneeth VepakommaAbstract
Uniform diffusion language models (UDLMs) generate text by iteratively denoising sequences of ordinary vocabulary tokens, any of which may be corrupted and subsequently revised. During training, UDLMs must denoise from a context that may itself be unreliable, whereas masked diffusion language models (MDLMs) always condition on a clean context. This challenging training objective makes UDLMs more FLOP-intensive to train than MDLMs, but it yields stronger downstream reasoning performance. Recently, selective looping, which recursively applies a few early Transformer layers within each denoising step, has been shown to improve training efficiency in MDLMs under an iso-FLOP budget. We investigate looping in UDLMs while keeping the training objective and sampler unchanged. We find that looping significantly improves training efficiency: looped UDLMs reach the validation NLL of a non-looped UDLM trained on about 0.5 trillion tokens with up to 2.8x fewer training FLOPs on OpenWebText and 3.3x fewer on FineWeb-Edu. We also find that, at a matched NFE budget, looped UDLMs reach up to 9.8% lower generation perplexity than the non-looped UDLM, with most gains achieved at moderate loop counts. Finally, we study different loop-count distributions and find that sampling smaller loop counts more frequently during training improves inference efficiency: on downstream tasks, this strategy uses 12% fewer loops than uniform loop-count sampling without degrading accuracy.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30066
PDF
-
What the Author Wrote, What the Agent Reads: Silent Cue Loss in Enterprise Document Ingestion
2026-10-21 16:51:25Yikai Zhao, Saurabh Pandey, Pradeep Kumar MisraAbstract
Enterprise agents act on documents. Before an agent can follow a policy, answer a compliance question or file a ticket, some ingestion layer must turn a Word file or PDF into tokens - and that layer is where an agent deployed in the wild silently loses information. We show that one layer is lost by every mainstream converter - the meaning an author encodes in how text looks - and that the loss makes agents confidently wrong rather than visibly incomplete. We build a task from 76 real policy documents: recover the requirements the author singled out, scored against the file's own OOXML properties by a deterministic matcher with no LLM judge. Formatting-blind ingestion collapses from micro F1 0.690 to 0.150 (docling) and 0.075 (plain text), a gap that replicates across six answering models from two vendors - five Claude models spanning the Opus, Sonnet and Haiku tiers, and OpenAI's GPT-5.6. An agent that cannot see the cue guesses, and docling's false-positive rate on requirements the author did not flag is four times ours. A span-level leave-one-out localises the loss to two cues - highlight (+0.302) and colour (+0.262), both Holm-adjusted p < 1e-4 - while bold, italic, monospace and strikethrough each sit within +/-0.003 of zero. The same asymmetry appears in PDF. Colour is also the cue no mainstream converter preserves and no parsing benchmark scores, and it is common: 32.8% of 6,460 documents carry a deliberate colour or highlight, 10.3% on an obligation. A pre-registered test shows that typing a cue for the agent adds +0.085 beyond merely preserving it (Holm-adjusted p = 0.049). We validate the labels against rendered page images through an independent modality (kappa = 0.72) and bound what wording alone can achieve (AUC 0.664). For agent builders the practical finding is that ingestion, not reasoning, is where this reliability is won or lost, and that the fix is cheap and measurable.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30026
PDF
-
Agent Behavior Maps: Locating Agents by What They Would Do, Without Running Them
2026-10-21 16:44:02Shota Fujikawa, Takuya Tamura, Kosuke Akimoto, Daichi Haraguchi, Masafumi OyamadaAbstract
LLM agents with similar success rates can follow different action sequences. An Agent Behavior Map represents LLM-harness combinations using centered action log-likelihoods on shared reference trajectories, without requiring new environment rollouts from the target agent. We derive error bounds relating squared coordinate distances to trajectory KL under explicit assumptions, and across eight cross-scored models the distances track the directly measured trajectory KL in rank order. Across 100 LLMs and four environments, ALFWorld-based coordinates support performance prediction and identify behavioral similarities. In ALFWorld, these similarities persist after controlling for success rates and model size, including across model families. As an extension, harness-component displacements approximately compose, with accuracy varying across components and environments.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30014
PDF
-
Building Auditable AI Text Detectors with Sparse Autoencoders
2026-10-21 16:41:14Liam Dugan, Callum McDougall, Matthew Denton, Philip Pham, Christine Kaeser-Chen, Chris Callison-BurchAbstract
AI-generated text detectors are increasingly deployed in consequential settings---peer review, education, admissions, journalism, and content moderation---where their outputs directly affect individuals. Yet the strongest detectors are fine-tuned neural classifiers whose decisions are opaque, while transparent alternatives (stylometric and linguistic-feature classifiers) underperform them by wide margins. We present a pipeline that converts a fine-tuned neural detector into a classifier that operates on human-readable concepts. We accomplish this by training sparse autoencoders (SAEs) directly on a detector's residual stream, assigning auto-interpretation labels to each SAE feature, and retaining only features whose generated descriptions pass a global fidelity threshold. We then train a lightweight classifier on just the interpretable features. We find that our SAEs recover latent concepts that correspond to highly specific and predictive linguistic "tells" of AI-generated text. Classifiers built on these extracted features are not only highly interpretable but exhibit stronger performance on out-of-distribution texts than either neural or stylometric alternatives.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30006
PDF
-
HiddenLens: A Multi-Layer LLM Framework for Detecting and Resolving Hidden Assumptions in Software Requirements
2026-10-21 16:35:35Tanu Singh, Aashna Suman, Vinod PatidarAbstract
Natural-language software requirements often omit assumptions about performance, operating conditions, user behaviour, and domain constraints, affecting architecture, testing, and deployment. If undiscovered until implementation, resolving them is far more costly than addressing them early. Although ambiguity and vagueness are widely studied, less attention has been given to assumptions embedded in requirements that appear clear and reasonable. We propose HiddenLens, a novel six-layer BERT-LLM pipeline and, to our knowledge, the first framework engineered to detect, classify, question, and resolve hidden assumptions rather than merely flag ambiguity. Layer 1 uses BERT attention entropy to localise unresolved spans; Layer 2 uses CoT-prompted LLM reasoning to articulate each assumption and its failure risk; Layer 3 uses fine-tuned BERT to classify assumptions into six non-functional requirement categories: performance, security, scalability, usability, reliability, and domain; Layer 4 generates ranked clarification options using FS LLM prompting; Layer 5 enables human validation through confirmation, editing, or rephrasing; and Layer 6 rewrites the requirement using the confirmed resolution, producing an explicit, testable, and traceable requirement. We also introduce the HiddenLens Corpus (HLC), comprising 312 requirements annotated with 847 hidden assumptions across six NFR categories. Inspired by PROMISE and PURE, HLC addresses the lack of annotated data for this task and is, to our knowledge, the first dataset developed specifically for it. HiddenLens outperforms keyword-based and QuARS-Adapted baselines by 71% and 61%, respectively, achieving an end-to-end F1-score of 0.79; its BERT-based NFR classifier achieves an F1-score of 0.847. These results indicate that HiddenLens can help analysts resolve hidden assumptions before they propagate through software development.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29996
PDF
-
Remembering Through Change: Provenance-Preserving Hypergraph Memory for Multimodal Agents
2026-10-21 16:34:28Jielong Tang, Ling-An Zeng, Xujie Yuan, Xiao Dong, Zhiqian Xia, Qingfeng Li, Shixiong Kai, Kaichao Liang, Rong Yao, Jian YinAbstract
Long-term memory for multimodal agents remains challenging, as it must preserve fine-grained visual evidence while tracking entity evolution across sessions. Existing abstraction-based memories compress images into text, often discarding visually discriminative details, whereas pixel-preserving memories retain raw observations but struggle to reliably associate diverse observations with persistent entity identities and track their evolving states. To address these limitations, we propose MHG-Mem, a multimodal hypergraph memory that jointly organizes entity identity, state evolution, and provenance to visual evidence. Specifically, MHG-Mem represents distinct entity mentions and their observed facts as individual nodes, grouping identity-consistent mentions and facts through occurrence hyperedges. It then synthesizes aspect-specific evolution states from these grouped observations and links them to supporting entities, facts, and source evidence through evolutionary and evidential hyperedges. To navigate this structured memory, we introduce hypergraph-based agentic search, which adaptively selects an entry layer, traverses typed hyperedges across memory levels, and inspects source observations when fine-grained verification is required. These designs support both detailed visual verification and reasoning over cross-session state changes. Extensive experiments across three benchmarks demonstrate the effectiveness of MHG-Mem, improving performance by 8.20-9.98% over the strongest baseline.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29995
PDF
-
CulturalRiddles: A Multicultural Benchmark of Riddles
2026-10-21 16:20:30My Chiffon Nguyen, Tianyi Hu, Vinod Anbalagan, Henry Gagnier, Pouya Sadeghi, Farah Abdou, Tidor-Vlad Pricope, Diego MMS Oliveira, Jocelyn K D'Arcy, Madeline Smith, Oluwaseun A Ajayi, Yusser Al Ghussin, Zena Al-Khalili, Micol Altomare, Shayan Amir, Ahanaf Aziz, Bryan Chen, Ruchira Dhar, Abdellah El Mekki, Rawan Mohammed Elframawy, Akhil Eppa, Silvia Fernandez, Alex Gimei, Arya Hariharan, Christian Hoang, RamMohanRao Kadiyala, Diana Kozachek, Peerat Limkonchotiwat, Ankita Maity, Di Mo, Ayush Munot, Aylin Naebzadeh, Anastasiia Nikitina, Dina Pisarevskaya, Divas Poudel, Jebish Purbey, Md Rabius Sany Apu, Ahmad Mufli Ramadhan, Ralitza Nikolova Rizova, Diogo M. Serrano-Miranda, Vivek Gopalkrishna Silimkhan, Luis Fernando Solis Navarro, Kaiser Sun, Arya Suneesh, Manaswitha Sunkara, Tejaswi Sunkara, Jonathan E Swindell, Hanna Yukhymenko, Jingyi Zheng, Esther Adenuga, Ilia Afanasev, Wafa Aissa, Lisa Alazraki, Lauren C Altomare, Priscilla Amondi Amuok, Emile Anand, Gabriel Assis, Jos\'{e} Mat\'{i}as Avalos, Silvia Lorena Avila, Zakaria Baannou, Niyati Bafna, Yash Bagla, Sarah Bahrami, Bontu F Balcha, Vanya Bannihatti Kumar, Irene Baucells, Abdumutalib Begmuratov, Naman Bhatia, Souvik Bhattacharyya, Isabel Blanco Ferreiro, Felipe Bochini, Sayuru Bopitiya, Angana Borah, Tulika Bose, Olena Bugaiova, Juliana Carvalho, Nancy F Chen, Stephy Chen, Rochelle Choenni, Jayanth Krishna Chundru, Leandro A Cossio, Neboj\v{s}a Damnjanovi\'{c}, Ona de Gibert, Marco De Santis, Pedro Delgado, Kaustubh Dhol\'{e}, Adnan El Assadi, Wesam El-Sayed, Nadine M Elframawy, Oumayma Essarhi, Naome A Etori, Reuben Chagas Fernandes, Saanidhi Pravin Gade, Javier Garc\'{i}a Gilabert, \'{A}lvaro Gonz\'{a}lez-Escapa, Bin Hu, Kaiyu Huang, Abderrahmane Issam, Danka Joki\'{c}, Ulugbek Khalilov, Dominik Krzemi\'{n}ski, Roy Ka-Wei Lee, Haichang Li, Chen Cecilia Liu, Zhengyuan Liu, Cedric Lothritz, Houman Mehrafarin, Antonio Miranda-Escalada, Fengran Mo, Mehrnaz Mofakhami, Wafaa Mohammed, Felipe Moreno-Vera, Prahitha Movva, Jannatul Nayem, Van Ngo, Joy Olusanya, Marc P\`{a}mies, Avinash Kuney Pandey, Yurii Paniv, Fred Philippy, Vladislav Poritski, Dikshya Poudel, Joanna Rado{\l}a, Sara Rajaee, Swati Rajwal, Prerana Rane, Muhammad Ravi Shulthan Habibi, Lucas Resck, Shubhashis Roy Dipta, Ananya Sahu, Eduardo S\'{a}nchez, Borja S\'{a}nchez Clemente, Aleix Sant, Nobin Sarwar, Istiak Shihab, Santhoshini Pranathi Singaraju, Jane Smith, David Stap, Emily Z Su, Ananth K Suresh, Sajag Swami, Ruiyang Tan, Nazia Tasnim, Khalid Tounoussi, Patteera Triamamornwooth, Muhammad Saad Umar, Shankar Venkitachalam, Oksana Volchek, Lingbo Wang, Debela Desalegn Yadeta, Rushikesh Zawar, Jinghan Zhang, Mike Zhang, Weihua Zheng, Marzieh Fadaee, Akhil Arora, K\'{a}roly Boczka, Julia KreutzerAbstract
As LLMs expand their language coverage, evaluating how well they understand and represent diverse cultures becomes increasingly important. However, many benchmarks remain English-centric, with limited cultural coverage and a focus on factual knowledge over its application in real-world cultural contexts. We propose CulturalRiddles, a benchmark of original, culturally grounded riddles created entirely by humans through a participatory community effort, spanning 62 cultures (45 countries, 51 languages). These riddles are deeply rooted in their respective cultures and require abductive and compositional reasoning, alongside detailed factual and social knowledge, language skills, and an understanding of culturally specific associations. We benchmark 28 LLM configurations using both automatic and large-scale human evaluation. Even frontier models achieve at most 22% accuracy in some communities. We publicly release all artifacts alongside an open contribution process to expand the benchmark's linguistic and cultural coverage.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29968
PDF
-
Commit or Abstain? System One Decision Models, LLMs, and Missing Legal Authority
Abstract
Given the enormous number of claims to be checked, verifiable legal AI requires scalable, cost-effective models that answer the practical needs of law. System One decision models are built on large language models but, instead of generating text, return a calibrated probability for each option. This suits the verification of legal authority, where the question is closed: a given holding, or none of them. We test whether such models abstain when the correct authority is missing. We extend CaseHOLD with an abstention condition in which the correct holding is replaced in 1,670 of 5,314 items, and compare four decision models with six generative LLMs. When no option is correct, generative LLMs still pick one in 38-86% of items; the most cautious decision model does so in 30%. On standard CaseHOLD, a decision model matches the best generative LLM (88.1% vs. 88.2%) at 1/55 of the cost. Accepting that LLM's answer only when a decision model agrees halves its wrong citations, and distilling a decision model's probabilities into Qwen3-8B, without labels, yields a well-calibrated local model. Legal AI should be judged by its wrong citations, not by accuracy alone.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29961
PDF
-
Towards English - Mizo Legal Machine Translation: Corpus Creation, Analysis, and Benchmarks
2026-10-21 16:11:01Telem Joyson Singh, Sanabam Boyshana, Sanasam Ranbir Singh, Priyankoo SarmahAbstract
Indian court proceedings are held in English, so speakers of languages with limited translation technology have limited access to them. We present the first English Mizo parallel dataset of Indian court judgments. It contains 13,168 sentence pairs from the Gauhati High Court and the Supreme Court of India. Professional native Mizo translators translated complete documents; sentence pairs were manually aligned, and two independent reviewers checked the results. Our study finds that English words are borrowed in a very selective way: official and legal terms are almost always kept in English, but words for everyday things like actions, situations, and results are usually expressed in native Mizo. Mizo diacritics are also used inconsistently across the corpus. When we fine tune the NLLB 200 1.3B model on our training data, it scores 27.47 BLEU on a balanced legal test set of 1,000 sentence pairs, an improvement of 7.16 BLEU over the base model. However, it still uses English words about three times more often than a human translator does.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29943
PDF
-
ToM2Vec: Continuous Theory-of-Mind Tokens for Multi-Agent Coordination
2026-10-21 16:05:54Nithin Sivakumaran, Ali Payani, Hyunji Lee, Mohit Bansal, Elias Stengel-EskinAbstract
Forming an accurate theory-of-mind (ToM) is key to successful multi-agent collaboration and requires reasoning under uncertainty to capture multiple possible partner states. This is hard to do in the text representation space native to large language model (LLM) agents, which more easily describe a single partner state. To address this challenge, we introduce ToM2Vec, which learns to form abstract theories-of-mind in a rich latent space via a fixed budget of continuous reasoning tokens. During training, supervision for these tokens comes from the partner's actual thoughts, which the ToM tokens are trained to abstract and match. At test-time, ToM tokens are generated by each agent and act as conditioners for that agent's thoughts and actions. Across several environments covering collaborative, competitive, and mixed-motive settings (Overcooked-AI, multiparty negotiation, Decrypto, Hanabi, and Collab Escape/Capture), we evaluate ToM2Vec's ability to support multi-agent interactions and collaboration, finding consistent improvements over the base model and a variety of competitive trained and prompted baselines. We show that latent reasoning is critical, with performance dropping when reasoning is expressed in text. Moreover, we show that ToM2Vec training generalizes to partner model families not seen at train-time, including a proprietary model (GPT-5.4). In our analysis, we provide evidence that the latent tokens encode a ToM, rather than task-specific information: ToM2Vec's tokens improve the agent's ability to directly answer questions about partner beliefs, desires, and next actions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29934
PDF
-
Graph Topology Meets Energy-Driven Pathfinding: Preserving Contextual Consistency in Retrieval-Augmented Generation
2026-10-21 16:05:44Yufeng Li, Zaijing Li, Fuchen Zhang, Zongzong Wu, Ming Zhao, Fengxiao TangAbstract
Retrieval-augmented generation (RAG) frequently integrates knowledge graphs (KGs) for complex multi-hop question answering. However, their performance is increasingly bottlenecked by the knowledge structuring and reasoning processes. (i) Severe semantic loss: Distilling continuous texts into discrete symbols strips away contextual nuances and narrative coherence. (ii) Pronounced modality mismatch: Fragmented graph structures clash with the continuous texts LLMs are pre-trained on, while rewriting them into natural language introduces generative noise and omits information. To address these limitations, we propose ThreadRAG, a semantic-topology-based RAG framework for context-preserving knowledge structuring and modality-native evidence reasoning. ThreadRAG introduces a Semantic Topology Mapper (STM), which organizes retrieved documents into context-intact Semantic Nodes and constructs a Semantic Topology Graph (STG) through narrative, semantic, and lexical edges. It further incorporates Energy-Driven Pathfinding (EDP), which identifies query-relevant subgraphs via energy diffusion and deterministically linearizes them into coherent textual evidence paths. Experiments on three multi-hop QA benchmarks demonstrate that ThreadRAG consistently outperforms strong baselines. Notably, it achieves a 14.4% EM improvement over text-centric methods on HotpotQA and surpasses the best baseline by 5.8% EM on 2WikiMultihop, while reducing token cost to 23.3% of the strongest baseline.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29933
PDF
-
Your Benchmark Is an Empirical Measure Over Difficulty
2026-10-21 16:02:37Yifan Sun, Naicheng Yu, Jiawen Gong, Jingyan Shen, Huan ZhangAbstract
A benchmark is not merely a collection of test questions; it naturally induces an empirical measure over difficulty. In this work, we develop this measure-based view of benchmark evaluation. This perspective allows us to identify and formalize two central failure modes: measure mismatch, where the benchmark difficulty distribution differs from the intended difficulty distribution, and coverage gaps, where uncovered regions along the difficulty axis prevent benchmarks from detecting meaningful localized training progress. We propose two diagnostic metrics that capture these two failure modes and provide rigorous theoretical guarantees. Empirically, we analyze 25 popular benchmarks spanning 7 domains using carefully estimated item difficulties. Our results show that many benchmarks implicitly overweight certain difficulty regions, so their leaderboard conclusions should be interpreted with caution; others leave substantial portions of the difficulty axis uncovered, limiting their ability to monitor the model improvement process. Finally, we turn diagnosis into action by introducing BenchFill, a simple difficulty-targeted item rewriting algorithm that effectively reduces local coverage gaps.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29927
PDF
-
Task-Preserving Multilingual Instruction Tuning for Open-Ended QA in WMT 2026 MIST
Abstract
This paper describes our submission to the WMT 2026 Multilingual Instruction Shared Task. We adapt Qwen3-8B with LoRA using organiser-provided data, EuroAlpaca, a multilingual instruction dataset created through task-preserving localisation, and CrossEuroAlpaca, a cross-lingual augmentation that assigns different source and target languages to instructions, contexts and responses. Our system targets open-ended question answering (QA-OEG) in ten shared-task languages represented in the auxiliary data. Under matched decoding, adaptation improves the mean Gemma-4 judge score on target-language QA-OEG by 7.8% relative to the base model. On an additional evaluation using the Aya Evaluation Suite across all shared-task languages, it yields relative gains of 41.5% in macro-averaged ROUGE-L and 2.9% in BERTScore-F1, with larger improvements in target languages. However, these gains are task- and language-specific. Gemma-4 scores improve for QA-OEG and summarisation in the target languages but decline for context-grounded QA there and across all three tasks in the non-target languages.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29926
PDF
-
Unintended Side Effects of Value Training in LLMs
2026-10-21 15:58:53Siddhesh Pawar, Chen Cecilia Liu, Anna Korhonen, Isabelle AugensteinAbstract
As LLMs are increasingly deployed in value-sensitive contexts, aligning them with human values has emerged as a post-training objective in NLP, aimed at instilling additional values and guiding model behaviour. However, current evaluations remain narrow. Because values are interconnected, it is unclear how fine-tuning for specific values alters unrelated capabilities and behaviours. In this work, we adapt four instruction-tuned LLMs across individual Hofstede's value dimensions and evaluate them on both cross-value transfer and broader behaviours. We find that value-specific training produces unintended side effects on untrained values, risk preferences, social cognition, and safety traits. Consequently, we suggest that value alignment after post-training should incorporate broader behavioral evaluations, and we highlight the need to investigate where additional value training might best fit into the pipeline.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29922
PDF
-
Benchmarking Large Language Models for Detecting Satirical Code-Mixed Sentences
2026-10-21 15:48:28R\u{a}zvan-Alexandru Sm\u{a}du, Ioana Buhnila, Daniela Tanase, Dumitru-Clementin Cercel, Florin Pop, Mihaela-Claudia CercelAbstract
Code-switching is a natural linguistic phenomenon, yet its impact on high-level reasoning tasks remains largely unexplored in under-resourced languages like Romanian. Natural language processing tasks, such as satire detection, rely heavily on nuanced cultural and syntactic cues that large language models struggle to capture. In this paper, we investigate four methods (i.e., random replacement, phrase translation, dependency parsing, and large language models) for generating synthetic code-mixed Romanian sentences with English, German, French, and Italian. We start with an existing dataset for satire detection, compiled from Romanian news articles and annotated at the sentence level by humans, and augment it with over 600k samples, namely CS-SeLeRoSa. We provide an in-depth analysis of the dataset generation process. Then, we evaluate several multilingual large language models across zero-shot, chain-of-thought, and fine-tuning regimes on both the synthetic corpus and a human-adjudicated gold benchmark, namely CS-SeLeRoSa-Gold. While older open models suffer substantial degradation (dropping to 37%-55% macro-F1), newer models like Gemma 4 31B achieve 68.1% zero-shot, rivaling frontier models like GPT-5 mini (69.5%). Furthermore, fine-tuning significantly reduces formatting errors and improves the results, increasing Gemma 4 31B to 77.5% on CS-SeLeRoSa and 72.6% on the gold standard, outperforming frontier zero-shot baselines.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29907
PDF
-
Investigating the Impact of Learner Errors on Social Bias in Large Language Models
2026-10-21 15:38:13Federico Marcuzzi, Martin Christoph, Iryna Gurevych, Alla RozovskaysAbstract
Existing social bias benchmarks evaluate large language models mostly on fluent, well-formed English, reflecting native speaker usage and leaving non-native writers unrepresented. To address this gap, we investigate how grammatical errors by non-native writers affect social bias. We develop a controlled framework that injects common learner errors into prompts and apply it to three bias benchmarks covering stereotype alignment and demographic preference across gender, race, and religion. We find that grammatical errors increase bias, and the effect grows with the error rate. Among the four common error types we consider, preposition errors produce the largest and most consistent increase in stereotype alignment. Our analysis suggests that errors increase bias when they are more likely to interfere with comprehension. Finally, fine-tuning on erroneous texts reduces the bias increase caused by grammatical errors. Our findings show that bias evaluations based exclusively on well-formed prompts may underestimate the bias experienced by users whose English contains grammatical errors.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29896
PDF
-
Acharya 2.0: An AI Tutor for Self-Learning Subjects in Hindi
2026-10-21 15:37:24Shantipriya Parida, Debasish Dhal, Sonal Khosla, Sambit Sekhar, Muhammed Abdur Rahmaan KamaldeenAbstract
Text processing is crucial for low-resource languages as it enables the development of tools and technologies that bridge the digital divide and provide equitable access to information and services. By empowering low-resource languages with robust text processing, we can preserve linguistic diversity, enhance education, improve communi- cation, and foster inclusivity in a rapidly digitizing world. Due to the lack of sufficient annotated datasets and lin- guistic resources, it is challenging to develop accurate and reliable language technologies for these low-resource languages. Large Language Models (LLMs) have attracted sig- nificant attention with their performance and the value they produce. However, with the ever-evolving language, adapting LLMs for a specific task is time-consuming and resource-intensive. Thus, Retrieval-Augmented Genera- tion (RAG) has overcome this limitation by augmenting LLMs with external knowledge, thereby improving their efficiency and extensive task-specific adaptation. Most applications of the RAG technique have been limited to English and has seen limited application in In- dic languages. This paper introduces the RAG-powered AI-Tutor application, Acharya 2.0, designed to produce academic assistance to Hindi-speaking students. The client-server-based web application integrates many fea- tures to support students in accessing and interacting with educational content.
Abstract (translated)
URL
https://arxiv.org/abs/2610.25192
PDF
-
Every Part Has Its Role: Part-of-Speech Effects in LLM Post-Training
Abstract
Large language models are typically trained with next-token prediction (NTP), which rewards predicting specific tokens rather than explicitly accounting for the semantic similarities between different words or expressions. Prior work has addressed this limitation by augmenting training data with minimally varying sentences that replace a target word with conceptually related alternatives, e.g., "The tired coach ran home" $\rightarrow$ {"The tired trainer ran home", "The tired instructor ran home", ...}. Such concept augmentation improves performance, but it remains unclear whether its effectiveness depends on where in the sentence the concept is expressed. We reproduce and extend prior work $\textit{beyond nouns}$ to verbs and adjectives, asking whether the part of speech (POS) carrying the concept signal affects perplexity and downstream performance. We find different patterns across evaluation settings: concept augmentation improves perplexity most consistently on news, independent of the POS used, whereas verbs provide the most reliable downstream gains. Despite these behavioral differences, representational analyses using CKA and linear probing show that the conditions yielding the largest downstream gains are not necessarily those that most alter the model's internal representations. Together, these results suggest that the effectiveness of concept augmentation depends on where and how conceptual information is introduced during training.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29843
PDF
-
Exploring Post-Training Alignment of Small Language Models for Biomedical Data-to-Text Generation: A Case Study of Medication Leaflet
2026-10-21 15:03:37Xi Yang, Guodong Liu, Chuqin Li, Fan Wu, Ergin Soysal, Min Jiang, Xing He, Jiang Bian, Yi Guo, Shams Zaman, Thomas Fuchs, Yonghui WuAbstract
Translating structured biomedical information into accurate, patient-friendly narratives remains a challenging biomedical data-to-text (D2T) task. While small language models (SLMs) are attractive for healthcare applications because of their lower computational cost, easier deployment, and improved data governance, the impact of post-training alignment on biomedical D2T generation has not been systematically studied. In this work, we compare four alignment methods: supervised fine-tuning (SFT), direct preference optimization (DPO), odds ratio preference optimization (ORPO), and group relative policy optimization (GRPO), using Qwen-2.5 models for medication leaflet generation. We further introduce FDALeaflets, a dataset derived from OpenFDA drug labels, to evaluate cross-regulatory generalization in a zero-shot setting. Selected aligned SLMs outperform GPT-5 on several reference-based metrics, whereas GPT-5 achieves better hallucination and adequacy scores on FDALeaflets under LLM-as-judge evaluation. Among RL-free alignment methods, ORPO achieves the best in-domain semantic performance, reaching a semantic textual similarity (STS) score of 0.9752. GRPO achieves the highest FDALeaflets STS (0.9407), exceeding GPT-5 (0.9295). These findings show that post-training alignment is a key factor in improving both semantic fidelity and generalization for biomedical D2T generation, while clinical safety and patient readability require human verification.
Abstract (translated)
URL
https://arxiv.org/abs/2607.13430
PDF
-
S-EDL: Eliciting Self-Evidence from Sequence Likelihoods for Semantic Calibration of LLMs
2026-10-21 14:55:58Yawei Li, Jiazheng Li, David R\"ugamer, Bernd Bischl, Yulan He, Mina RezaeiAbstract
Calibrating large language models (LLMs) in open-ended generation is uniquely challenging because many distinct token sequences can express the same underlying meaning. Prior calibration techniques target either fixed-set classification confidence or the binary correctness of individual strings, making them inherently ill-suited for the dynamic, multi-string nature of open-ended semantic outcomes. To bridge this gap, we introduce S-EDL, a novel evidential method that natively optimizes LLMs for semantic calibration. By eliciting evidence from the sequence likelihoods of the model itself, S-EDL constructs a differentiable Dirichlet prior over prompt-specific semantic classes. Optimizing the evidential loss under this prior yields a calibration-aware training signal over variable semantic classes, completely bypassing the need for fixed multiple-choice options. Evaluations across four open-ended benchmarks and three models demonstrate that S-EDL substantially reduces semantic calibration error while improving accuracy. By aligning generative likelihoods with semantic correctness, this work establishes a framework for deploying trustworthy LLMs in complex, open-ended applications. Code is available at: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29832
PDF
