Leaderboards, eval harnesses and ablations — the contested business of deciding which model is actually better.
arXiv:2609.39972v1 Announce Type: new Abstract: Speculative decoding accelerates language model inference by verifying multiple draft tokens in a single target-model pass. Recent parallel drafters ha...
arXiv:2609.39975v1 Announce Type: cross Abstract: This paper presents an overview of the fourteenth edition of the BioASQ challenge, organized in the context of the Conference and Labs of the Evaluat...
arXiv:2609.40097v1 Announce Type: new Abstract: Existing auto-research benchmarks often entangle multiple sources of improvement, including training frameworks, hyperparameters, compute budgets, and...
arXiv:2609.40124v1 Announce Type: new Abstract: Bias has long been studied in social psychology and cognitive science, where decades of research have produced a body of validated interventions that r...
arXiv:2609.40181v1 Announce Type: new Abstract: We introduce Index-Translate, a multilingual translation model family that combines a shared multilingual foundation with specialized training for gene...
The paper introduces the problem of cross‑lingual loopholes in large language model (LLM) unlearning, where forgetting a fact in one language can leave it accessible in others. It presents a new 174‑language benchmark, the Cross‑Lingual Unlearning Tensor, and proposes COVER, a method that selects a subset of source languages to maximize unlearning coverage under a language budget. Experiments show COVER reduces residual knowledge by 7.8–27.3% compared to uniform selection and works on both synthetic and real low‑resource news data.
arXiv:2609.38324v1 Announce Type: cross Abstract: Multi-agent systems of LLMs add discussion to majority voting and are therefore expected to be more capable. However, empirical reports conflict on w...
arXiv:2609.38426v1 Announce Type: cross Abstract: We introduce LoopVL to study whether Loop Transformers can be effectively extended to vision- language models. LoopVL combines Module-Loop and Model-...
SecureVibe is a training recipe designed to enhance the security of vibe coding by explicitly targeting planning and testing for code security. It combines supervised fine‑tuning on a security suite with post‑training methods (SECUREVIBE_rl and SECUREVIBE_hg) that use verifiable execution feedback and hint‑based self‑supervision. The approach outperforms baselines on multiple security coding benchmarks, improving security pass@1 by up to 6.9 points on BaxBench and 11.5 points on SusVibes, while also boosting functionality pass@1 on both security and generic coding tasks.
arXiv:2609.39334v1 Announce Type: cross Abstract: Test-time scaling has recently emerged as a powerful approach for improving LLM reasoning by allocating additional computation during inference, subs...
arXiv:2609.39688v1 Announce Type: cross Abstract: Multimodal encoders such as CLIP underlie many downstream systems, but their web-scale training data embed harmful associations that safety alignment...
The paper introduces FIGS, a dual‑axis evaluation framework for multi‑turn sycophancy that avoids penalizing empathy. It uses a 10‑turn conversational simulator with 500 diverse scenarios to test whether models stay truthful while keeping praise proportional, and whether they show calibrated validation of user feelings. The study finds that current models either drift toward sycophancy or become overly detached, highlighting an unresolved trade‑off in sustained dialogue.
arXiv:2609.39920v1 Announce Type: cross Abstract: Large vision-language models (LVLMs) exhibit strong multimodal in-context learning (ICL) capabilities, yet this ability degrades substantially as mod...
arXiv:2510.13935v3 Announce Type: replace-cross Abstract: The facts a language model stores are tied to its parameter count, so small models that fit on edge devices fail on expert problems, which ne...
arXiv:2603.18822v2 Announce Type: replace Abstract: What is known about the values of national populations rests almost entirely on questionnaires, which prompt respondents to rate researcher-supplie...
arXiv:2606.15911v2 Announce Type: replace Abstract: This paper focuses on automatically generating informative ad descriptions in sponsored search. Unlike ad titles which are usually optimized to att...
The paper evaluates safety monitors by measuring recall only on prompts that the target model actually answers, rather than on all harmful prompts. Across several guard systems, recall at a 1% false‑positive rate drops sharply when focusing on answered prompts, with monitors catching refused requests 1.1–6.4 times more often than answered ones. Rewriting prompts to be less explicit dramatically increases compliance and reveals that many harmful requests slip past monitors, especially when phrasing is softened. Fine‑tuning guards on these rewritten prompts improves recall from 0.24 to 0.89 on answered requests and generalizes to unseen benchmarks.
arXiv:2609.31688v2 Announce Type: replace Abstract: In verifiable domains such as math and coding, finding one correct solution among many attempts can matter more than the pass rate of each attempt....
arXiv:2609.34284v2 Announce Type: replace Abstract: Personalized LLMs must decide, for each stored preference, whether the current context calls for applying or suppressing it, which we call its appl...
The paper introduces LSR‑Ben, a benchmark designed to evaluate process reward models (PRMs) on scientific and logical reasoning tasks, addressing a gap left by existing math‑focused benchmarks. Experiments on 22 models reveal that PRMs and LLMs perform poorly in non‑mathematical domains, with LLMs tending to over‑identify errors while PRMs tend to overlook them. LSR‑Ben aims to spur research that broadens PRM applicability and improves LLM reasoning.