paperlist.ai

AI conference papers

AI conference paper library

Search papers from 30+ top conferences, including NeurIPS, ICLR, ICML, CVPR, ACL and AAAI. Filter by keyword, then open any paper for its abstract, authors and links.

arXiv

When Lexical Change Misleads: Rethinking Dynamic Topic Model Evaluation with Traditional and LLM-Based Metrics

Charu Karakkaparambil James

Dynamic topic models capture evolving word distributions, but traditional coherence metrics may fail when vocabulary changes while semantic meaning persists. We evaluate 120 topics from CoNTM and DLDA across NYT, DBLP, and arXiv, using three human annotators and Low, Medium, and High lexical-change categories. Traditional temporal coherence shows highly variable agreement with human judgments ($ρ$=-0.256 to 0.614). In contrast, LLM-based semantic similarity agrees strongly with human semantic judgments for CoNTM on NYT ($ρ$=0.609), DBLP ($ρ$=0.721), and arXiv ($ρ$=0.502), but is less consistent for DLDA. Lexical-change stratification reveals variation hidden by aggregate evaluation. We therefore advocate lexical-change-aware evaluation, jointly reporting traditional coherence and LLM-based semantic measures as complementary rather than interchangeable signals.

arXiv

VoiceChat-TTS: A Low-Latency Continuous Speech Synthesis Model for Interactive Agents

Edresson Casanova, Jaehyeon Kim, Mariana Graterol Fuenmayor, Shehzeen Hussain, Viacheslav Klimkov, Valentin Mendelev, Mikyas Desta, Paarth Neekhara, Piotr Zelasko, Chen Chen, Elena Rastorgueva, Ke Hu, Ankita Pasad, Xuesong Yang, Aya Alja'fari, Rajarshi Roy, Rohan Badlani, Jason Roche, Jason Li, Zhehuai Chen

Spoken dialogue is a natural form of human--computer interaction, yet most speech language models remain limited to turn-based operation and lack real-time adaptability, such as user barge-in. Recent duplex speech-to-speech and speech-to-text models reduce latency by replacing multi-stage pipelines, but often compromise speech quality because accurate ASR, interruption handling, and high-fidelity synthesis must be optimized jointly. We propose VoiceChat-TTS, a low-latency, continuous, and streamable text-to-speech model for interactive agents. VoiceChat-TTS is driven directly by LLM text-token streams, supports explicit interruption via control tokens, and produces silence when no textual input is available. The model enables always-on, responsive speech generation while preserving modularity and high speech quality, and it supports mid-utterance interruptions without resetting the KV cache.

arXiv

From Passive Delegates to Strategic Negotiators: Reinforcing Social Reasoning in Small Language Models with SocialRL

Wenyue Hua, Zachary Huang, Tyler Payne, Safoora Yousefi, Saleema Amershi, Asli Celikyilmaz

AI agents increasingly act on their users' behalf, handling tasks such as scheduling meetings, comparing offers, and haggling over prices. These principal-driven tasks routinely place the agent across from a counterpart (another user's agent, a seller, a recruiter) whose goals may conflict with its principal's. Yet the dispositions that make an assistant pleasant can make it a poor delegate: a friendly, helpful frontier model may disclose its principal's private information unprompted and concede at the first sign of resistance. We present SocialRL, a general recipe that trains social reasoning directly, and apply it to a 4B model across six domains: Deal-or-No-Deal, CaSiNo, Craigslist, Job Interview, Calendar, and Marketplace. Every domain is trained in-domain under the same recipe, and every policy is evaluated on all six. We find that (1) in-domain training reaches the frontier: on held-out scenarios the 4B matches or exceeds the GPT-5 family per domain, closing 73-122% of the baseline-to-frontier gap on the negotiation games, with 78% of buyer openings anchoring below target versus 3% untrained; (2) cross-domain transfer follows game structure: structurally paired games lift each other, a broad multi-issue donor lifts nearly all domains, and structurally isolated games transfer nothing; (3) guided by this transfer structure, two strategies, cascade RL and multi-teacher on-policy distillation (OPD), consolidate the per-domain specialists into a single unified 4B that reaches 0.627 average utility across all six environments, matching or exceeding GPT-4.1 (0.625), GPT-5.1 (0.619), and GPT-5.2 (0.613); (4) an explicit theory-of-mind scaffold helps only through training: distilling the ToM trace, rather than actions alone, lifts utility on every environment and generalizes better across them, and of the two ToM skills, only next-action prediction predicts negotiation outcomes.

arXiv

Do AI chatbots find what experts would? Effects of model, user role, and sample size on study retrieval for medical questions

Qingfang Liu, Qiao Jin, Joe D. Menke, Thorsten Kahnt, Zhiyong Lu

Large language model (LLM) chatbots are increasingly used to answer clinical questions with citations to relevant clinical studies. Prior research has largely focused on citation fabrication, leaving a gap in evaluating the quality of retrieved studies and the factors driving their selection. In this study, we evaluated three general-purpose LLM chatbots: Claude Sonnet 5, Gemini 3.1 Pro, and ChatGPT GPT-5.5. We prompted the models with clinical questions adapted from 20 review questions in Issues 6 and 7 of the 2026 Cochrane Database of Systematic Reviews, simulating patient, clinician, and evidence-synthesis researcher roles. Each chatbot was queried under each user role with four independent repetitions, yielding 720 responses. Each chatbot was asked to support its answers with primary clinical citations, which we benchmarked against the included and excluded study sets of the Cochrane reviews. On average, a chatbot response retrieved 39.2% $\pm$ 29.8% of Cochrane included studies, while citing 5.0% $\pm$ 9.4% of excluded studies. Recall of Cochrane included studies varied significantly by model and user role. ChatGPT achieved higher recall than Claude or Gemini (63.1% $\pm$ 29.5% vs. 37.0% $\pm$ 23.8% vs. 17.3% $\pm$ 13.1%; $p=2.0\times10^{-5}$). The researcher role yielded higher recall than the clinician or patient roles (42.8% $\pm$ 30.8% vs. 38.6% $\pm$ 28.9% vs. 36.1% $\pm$ 29.3%; $p=2.0\times10^{-5}$). Controlling for publication year, citations per year, and open-access status, sample size was the only independently significant predictor of retrieval (odds ratio 1.80 per 1-unit increase in log sample size, 95% CI 1.37-2.36, $p=2.34\times10^{-5}$). These findings suggest that while LLM chatbots can retrieve some studies identified by expert reviewers, their performance varies by model and user role, and they exhibit a bias toward clinical trials with larger sample sizes.

arXiv

Amplified Does Not Mean Predictive: Reasoning Behaviors in Thinking Models

Jean de Dieu Nyandwi, Leena Mathur, Yonatan Bisk, Robert Hawkins, Graham Neubig

Which reasoning behaviors are associated with correct answers in reasoning models, and does reasoning-oriented training amplify those behaviors? This distinction is important because reasoning-oriented training can make traces look more deliberative without amplifying the behaviors most tied to model correctness. We quantify this mismatch with Behavioral Lift, a metric that measures how much correctness changes when a behavior is present versus absent in a model's reasoning trace. Across 15 models and 6 benchmarks spanning text-only and vision-language reasoning, we annotate 15,282 traces with a taxonomy whose core behaviors are defined for both LLM and VLM traces. We find evidence for an Amplification-Lift Gap, in which thinking models strongly amplify self-correction, hypothesis testing, and uncertainty acknowledgment, while the highest-lift behaviors are confidence calibration, knowledge alignment, and self-awareness. Confidence calibration is among the strongest positive signals of correctness in both modalities, yet is barely amplified; uncertainty acknowledgment is amplified by 3--7$\times$, yet is weakly or negatively associated with correctness. We find that reasoning-oriented training does not preferentially amplify the highest-Lift behaviors, motivating process-level objectives that reward calibrated and grounded reasoning rather than surface form alone.

arXiv

BM25-Augmented Many-Shot Translation for Low-Resource North-Eastern Indian Languages

Aashish Dhawan, Christopher Driggers-Ellis, Dzmitry Kasinets, Christan Grant, Daisy Zhe Wang

This paper describes the University of Florida Gators submission to the WMT26 Low-Resource Indic Language Translation shared task. We adapt the retrieval-augmented many-shot translation pipeline from our AmericasNLP 2026 system to translate between English and eleven North-Eastern Indian languages in both directions. At inference time, BM25 retrieves the most similar parallel examples from a language-specific training bank, and Gemini 2.5 Flash translates the input conditioned on these examples. No model fine-tuning is involved. Training banks combine official WMT26 data with publicly available corpora such as Samanantar and prior WMT shared task releases. A grid search over retrieval count r and development exemplar count d across all 22 language-direction pairs selects the best configuration for each submission.

arXiv

Capacity-Dependent Effects of Data Selection for Reasoning

Cuong Dang, Hoang Anh Just, Ruoxi Jia

In reasoning supervised fine-tuning, candidate responses for the same instruction can differ substantially in how well they match the student's current distribution. Recent likelihood-based response selection methods suggest that responses closer to the student distribution provide more effective supervision, motivating the hypothesis that high-likelihood responses may generally be preferable for fine-tuning. In this paper, we revisit this intuition and show that the value of likelihood-based data selection depends critically on model capacity and training duration. Through controlled experiments on mathematical reasoning, using students ranging from 1.5B to 8B parameters and supervision generated by stronger teacher models, we observe a clear \emph{capacity-dependent} ``{\color{SMALLCOLOR}\textbf{Fast-Fit}} / {\color{LARGECOLOR}\textbf{Slow-Gain}}'' pattern. High-likelihood data provides faster and more stable early improvements, especially for smaller models, but low-likelihood data becomes increasingly beneficial for larger models when training is allowed to continue longer. To explain this phenomenon, we analyze learning dynamics, showing that small models often fail to absorb low-likelihood supervision and instead fall into shallow or repetitive behaviors, while larger models are better able to move toward the teacher distribution under such data. We further provide a capacity-constrained theoretical view of distillation that clarifies how data difficulty, data span, and student capacity jointly govern transfer. Overall, our findings show that effective data selection for reasoning should be aware of model capacity and computing budget rather than based on a single universal preference for high-likelihood supervision.

arXiv

StreamHear: Domain-Adapted Pseudo-Labeling for Semi-Supervised Streaming Speech Recognition

Zefang Liu, Chenyang Zhu, Sangwoo Cho, Xujun Peng, Shi-Xiong Zhang, Sambit Sahu

Streaming automatic speech recognition (ASR) underperforms on domain-shifted target audio, where labeled in-domain data is costly to prepare while unlabeled audio is abundant. We present StreamHear, a semi-supervised pipeline that adapts a pretrained streaming student by fine-tuning an offline transducer teacher on the labeled training set, generating pseudo-labels on the unlabeled portion, and fine-tuning the student on the mixture. We further introduce a prior-regularized dynamic-programming realignment step that fixes chunk-level word placement using an ASR-hypothesis anchor. Across four datasets spanning financial calls, prepared read speech, and phone-quality dialogue, StreamHear consistently outperforms supervised student fine-tuning and narrows the gap to the offline teacher.

arXiv

Reading Between The Lines: Modeling and Evaluating Behavioral Realism in Legal Simulation

Divya Vetticaden, Arya Gupta, Julian Nyarko, Megan Ma

Deposition training requires attorneys to manage dynamic witness behavior, yet legal-AI evaluations largely focus on factual accuracy, reasoning, or response-level plausibility. We introduce WitnessSim, a deposition simulator driven by controllable legal personas. We use an evaluation framework separating behavioral realism from pedagogical usefulness. We assess realism through adversarial testing, blinded attorney comparison, and analysis of longitudinal behavioral trajectories. WitnessSim generally maintained plausible behavioral boundaries, and attorneys did not systematically prefer either original testimony or WitnessSim generated testimony. Pedagogical tests showed that witness behavior changed meaningfully in response to question form and attorney intervention without uniformly collapsing the assigned persona. Together, these results showcase a model of behavioral fidelity in legal simulations, and provide a framework for evaluating its performance.

arXiv

TeachMateGPT: A Multi-Agent Knowledge-Grounded Framework for Pedagogical Assessment Generation from Science Curriculum Materials

Fatema Tuj Johora Faria, Mukaffi Bin Moin, M. F. Mridha, Jubayer Al Mahmud

Automatically generating textbook-grounded assessment items can reduce science teachers' workload, but existing retrieval-augmented generation (RAG) systems rely on flat retrieval, support only single-question generation, lack safeguards against weak evidence, and are ill-suited to low-resource, board-exam-structured curricula. We address these limitations with TeachMateGPT, a multi-agent system contributing four advances to curriculum-grounded science-assessment authoring. (i) COPE, a hierarchical knowledge base replacing token-window chunking with a multi-resolution index that segments documents along syllabus structure and links them at three granularities via a traversable graph-based lineage, matching evidence to each topic's instructional level. (ii) A staged, fail-closed agent pipeline replacing one-shot retrieve-then-generate: routing gates search, retrieval fuses dense and lexical evidence under a coverage gate that withholds generation on insufficient evidence, and specialist agents draft objective and constructed-response items. (iii) SAVER, a source-attributed verification protocol scoring faithfulness, relevance, and hallucination risk against retrieved evidence, applying stricter grounding checks across each creative question's four sub-parts, paired with teacher-in-the-loop evaluation rather than automatic filtering. (iv) NCTB-SciGen8, a curriculum-grounded dataset of 198 items (143 multiple-choice, 55 creative questions) spanning all 14 chapters of the NCTB Class 8 science textbook, produced by the pipeline and rated by three practicing teachers. TeachMateGPT raises faithfulness (0.68 $\rightarrow$ 0.96) and answer relevancy (0.60 $\rightarrow$ 0.89) over a vanilla RAG baseline.

arXiv

GRPO Beyond English: A Large-Scale Study of GRPO in Non-English and Multilingual Settings

Konstantin Dobler, Federico Scozzafava, Jonathan Janke, Mohamed Ali, Simon Lehnerer

Reinforcement Learning with Verifiable Rewards (RLVR), often optimized with Group Relative Policy Optimization (GRPO), has become a central recipe for improving the reasoning capabilities of pretrained language models but current studies remain heavily English-centric. We conduct a large-scale empirical study of multilingual and non-English GRPO across a wide range of base models, training languages, and different reasoning language rewards. We find that training to reason in the native language often leaves only a small gap to training for English reasoning. We further observe strong crosslingual transfer: training in one language often improves performance in many others. However, specific trends are highly model- and language-dependent. In some cases, training in a particular language induces severe regressions on out-of-domain capabilities in other languages. Our analysis shows that RLVR beyond English can provide broad crosslingual gains, but also requires broad evaluation to detect language-specific regressions.

arXiv

Asymmetric Discourse Homogenization and Shared Language Technology: Evidence from Reddit

Fengming Liu

I document an ideologically asymmetric break in the pre-existing diversification trend of political discourse, emerging around late 2022, using 6 million Reddit comments from two cross-partisan forums, 2019-2025. Conservative users experienced an interruption of their prior diversification trajectory; progressive users showed no comparable change. The asymmetry is consistent across estimation strategies (ITS, DiD, RDiT, propensity-score matching) and temporal aggregations. A daily-frequency permutation test over 2,377 candidate cutoff dates shows the ChatGPT threshold produces an unremarkable estimate (49.8th percentile): the shift builds gradually instead of breaking at a single date. A continuous cumulative LLM index, tracking AI exposure across seven model releases, remains significant under a quadratic trend specification that eliminates the binary estimate. A stayer analysis narrows the mechanism: the homogenization effect disappears when the sample is restricted to authors active throughout the study period, and the stayer confidence interval excludes within-author effects even a tenth the size of the full-sample estimate. The mechanism is most parsimoniously ecological (community-level discursive convergence) rather than individual-level AI adoption, though the data cannot cleanly separate this account from concurrent secular change.

arXiv

A Calibrated Test of Internal Action Maps: State Signals Without Global Affine Closure

Dekun Yang

A hidden state signal can be decodable or causally usable without supporting a reusable action map. We test whether action maps fitted without a source reach its natural post-action activation and compose. We organize the tests as an evidence lattice and validate the geometric branch on a known affine S_5 carrier: all held-source folds pass one-step, composition, inverse, decoding, and commutativity gates. Structured curvature and held-domain conjugacy raise error monotonically, but only 23/30 strongest cells flip a closure gate, bounding rather than universalizing calibration. In post-trained Qwen/Qwen3-4B, frozen final-token h28 affine maps have mean held-entity error .519, versus .398 for within-test-domain cross-fit. Seven randomized entity splits and map geometry do not support a purely entity-specific account. Earlier h4/h16 layers fit one-step transitions better, but h4 conflict-state decoding is weak and lexical controls remain unresolved. Three matched intervention datasets regenerated from one frozen checkpoint show causal effects only at h28/h36. Outcome-aware refitting improves h28 one-step error to .474 (.469 with weighting), yet no refit passes composition. Learned finite worlds likewise preserve relative algebraic signals or shared charts without held-source affine closure. Within the tested carriers, state availability, causal use, local geometry, and reusable closure are separable. The result is limited to one pretrained model, sampled final-token layers, two finite worlds, and the tested affine or diagnostic function classes.

arXiv

Measuring Fairness in Large Audio Language Models via Semantic-Aware Bias Estimation

Zhe Liu

Large Audio Language Models (LALMs) have seen increasing use for audio understanding tasks such as speech recognition and audio question answering, raising concerns about fairness across demographic subgroups. Fairness evaluation in spoken-input settings is challenging due to confounding factors, including semantic variation in spoken content and speaker-specific characteristics. Ignoring these factors can result in misleading conclusions about model bias. We propose a semantic-aware mixed-effects regression framework for fairness evaluation in LALMs that explicitly accounts for these confounders. Our approach incorporates sentence-level semantic embeddings of reference text as covariates and models speaker identity as a random effect. Notably, semantic representations are extracted from the same LALM under evaluation, enabling semantic control over variation as perceived by the model itself. Experiments on simulated data and real-world benchmarks demonstrate that the proposed approach substantially reduces spurious fairness findings and yields more robust and interpretable estimates of subgroup performance differences.

arXiv

ARC: Fair Relative Advantage Comparison in Open-Ended Real-World Interaction

Yongqi Tong, Tan Li Hui Faith, Choy Zhen Wen Marcus, Zhou Jin, Kewei Fu, Jiang-Ming Yang, Jianshe Li, Xin Zhang

Open-ended real-world interaction admits multiple valid behaviors: an agent may answer directly, ask for clarification, provide progress updates, or confirm before acting. This flexibility breaks a core assumption behind group-based RL: rollouts compared within a group are no longer guaranteed to be behaviorally comparable. As a result, reward-model preferences over interaction style can distort relative advantages and steer optimization toward reward-preferred behaviors rather than context-appropriate ones. We formalize this as a \textit{reward fairness problem} and propose \textbf{ARC} (Advantage Regularization via Conditioning), a training recipe that restores fairer relative comparison through strategy-conditioned rollout grouping, together with hybrid rewards and entropy regularization. We study ARC in our proposed \inter, a novel paradigm for responsive, steerable, and execution-aware user-agent interaction that decouples user-visible communication from latent reasoning and tool use. \inter\ also provides the annotation and distillation pipeline for constructing \inter-86K, our strategy-annotated training corpus for supervised and RL training. Empirically, ARC substantially strengthens the core $τ/τ^2$ tool-use benchmarks, while \inter\ reduces time-to-first-token from 4.91s to 1.27s relative to a think-style baseline. Together, these results suggest that a central bottleneck in open-ended interactive learning is not only how agents are rewarded, but whether their behaviors are compared fairly in the first place. The ARC implementation and \inter-86K training data will be released.

arXiv

Is Convergence Inevitable? Tracing Output Homogeneity Back to Base Models

Alexandrine Fortier, Hazel Chen, Peter West

The lack of diversity in LM content is widely attributed to the alignment process, but how and where exactly in the pipeline this collapse begins is unknown. We argue that output homogeneity is likely learned during the pretraining phase, and only revealed or magnified during the alignment process. Specifically, we find that semantic convergence is observed from the first alignment stage--the instruction-tuning phase (SFT)--suggesting that homogeneity might already exist in the pre-alignment model. To investigate this, we conduct controlled SFT experiments examining how training data influences output convergence on specific input/output pairs. We find that convergence can be revealed and amplified, but not introduced by the SFT data, supporting its role as a catalyst rather than a cause. To further test whether homogeneity originates before alignment, we measure convergence in base models. We find that instruct-like collapse can be induced through prompting alone, even without alignment. Taken together, our results suggest that semantic convergence may arise naturally from the objectives underlying LM training, making it difficult to mitigate through post-alignment interventions alone.

arXiv

HalluTruthQA-4K: A Fine-Grained Corpus and Annotation Process for Arabic Hallucination Detection and Truth Verification

Salah Eddine Bekhouche, Abdessalam Bouchekif, Hichem Telli, Mohammed-En-Nadhir Zighem, Abdenour Hadid

Large language models can generate fluent Arabic answers while introducing factual errors that are difficult to identify and verify. Existing Arabic hallucination resources often assign a binary label to an entire response, indicating whether it is hallucinated or non-hallucinated, but provide limited information about the exact erroneous content, the reason for the error, or the correct factual answer. We present HalluTruthQA-4K, an expanded version of the HalluTruthQA resource containing 4,000 expert-curated Arabic question-answering instances across four knowledge-intensive domains: Islamic knowledge, history, science, and geography. Serving as the official dataset for Track 2 of the HalluScoring 2026 shared task, HalluTruthQA-4K extends our original corpus to 4,000 instances. Each instance pairs an Arabic question with a model-generated response, a verified reference answer, and five plausible distractors. Hallucinated responses are additionally annotated with character-level erroneous spans, human-written explanations, and hierarchical hallucination types. The corpus contains 1,643 hallucinated and 2,357 non-hallucinated responses, with 1,843 annotated erroneous spans. We describe the resource construction and annotation methodology, including question selection, controlled answer generation, candidate construction, expert annotation, independent verification, adjudication, and quality control. We also document the annotation guidelines, taxonomy, data format, inter-annotator agreement, and corpus statistics. HalluTruthQA-4K provides a reusable resource for hallucination detection, span-level error localization, explanation generation, factual verification, and the broader evaluation of factual reliability in Arabic language models.

arXiv

Can AI Agents Simulate A/B Test Outcomes? A Validation Framework for Agentic Experimentation

Stefan Hut, Lorenzo Masoero

A/B testing remains the standard for rolling out new features in the technology industry. Each experiment, however, consumes real traffic, engineering effort, and weeks of wall-clock time. Can AI agents---conditioned on behavioral profiles and contextual descriptions of the intervention---simulate outcomes accurately enough to vet candidate treatments before committing live traffic? We formalize this question as a \emph{Simulated Randomized Controlled Trial} (S-RCT) and derive a two-layer error decomposition that separates agent approximation error from subsampling error, enabling targeted improvements to each. The framework is agent-agnostic: any behavioral model---from a fine-tuned specialist to a general-purpose foundation model---can serve as the simulation engine. Validated on 67 historical marketing A/B tests, a baseline S-RCT using an off-the-shelf foundation model captures directional signal (sign overlap 0.70) but systematically overshoots effect magnitudes. A two-phase pre-period calibration protocol reduces the squared prediction error (after removing irreducible measurement noise) by ${\sim}77\times$; a within-subject design---where each agent is exposed to both arms---reduces standard errors by ${\sim}2.4\times$. We discuss limitations of the current approach and identify applications where experimenters stand to benefit from agentic signals.

arXiv

Dissociating the Internal Representations of Sycophancy in LLMs

Anthony Baez, Sheer Karny, Pat Pataranutaporn

Large Language Models (LLMs) frequently exhibit sycophancy, agreeing with a user's statement even when it is incorrect. While often studied as a single, uniform behavior, sycophancy can manifest in substantially distinct ways across contexts, raising the question of whether this heterogeneity is reflected in its internal mechanisms. To address this gap, we dissociate the representations of sycophancy into factual and opinion subtypes, motivated by prior evidence of heterogeneous truth representations in LLMs. We train linear probes and construct steering vectors on one subtype's activations and evaluate their transfer to the other, measuring the extent to which representations are shared and visualizing them via Linear Discriminant Analysis. We find that different LLMs represent these subtypes differently, with either more aligned or more distinct representations, and apply this insight to improve representational interventions for reducing sycophancy. Our dissociation method offers a general framework for studying the representational structure of complex model behaviors.

arXiv

SomaliBench Eval: Measuring English-to-Somali Refusal Gaps in Open-Weight Language Models

Khalid Yusuf Dahir

Large language model safety evaluation remains heavily English-centered, leaving low-resource languages under-measured even when models are deployed globally. We evaluate four open-weight instruction-tuned models on SomaliBench v0, a native-author-verified benchmark of 100 harmful-intent prompts paired across English and Somali. Each of Llama-3.1-8B-Instruct, Gemma-2-9B-Instruct, Qwen-2.5-7B-Instruct, and Aya-23-8B is run locally with temperature 0 and the same English "helpful, harmless, and honest" (HHH) system prompt. We find large English-to-Somali refusal gaps for all four models, ranging from 0.40 to 0.93, all strictly positive under a paired bootstrap and significant by exact McNemar tests. For three models, the dominant Somali non-refusal mode is not fluent harmful compliance but unclear output: wrong-language, incoherent, or off-topic generations. A pinned Claude Sonnet snapshot (claude-sonnet-4-5-20250929) classifies each response as refused, complied, or unclear; its safety layer declined 34 of 800 classifications, which the native author labeled manually. A native-author spot-check achieves 100% agreement with the judge (Cohen's $κ=1.00$) on 74 comparable rows. We report aggregate refusal rates, category gaps, and reliability statistics only; raw model generations are retained locally and are not released because some may contain harmful content.

Keep browsing

More papers from top AI conferences

30+ top conferences, by conference, year, and topic.