Leaderboards, eval harnesses and ablations — the contested business of deciding which model is actually better.
arXiv:2610.26929v1 Announce Type: cross Abstract: Sanskrit verse recitation is governed by $\textit{Chandas}$, a strict prosodic framework defined by rhythmic $\textit{Laghu}$ (light) and $\textit{Gu...
arXiv:2610.27205v1 Announce Type: cross Abstract: Web agents are typically evaluated on English-language interfaces, leaving their ability to operate across the world's multilingual websites largely...
The paper introduces AuthorizationBench, a benchmark designed to evaluate whether large language model (LLM) agents act within the scope of user authorization when acting on their behalf. The benchmark comprises 286 executable tasks across ten application domains, covering situational triggers, presence or absence of direct harmful consequences, and varying reasoning effort. In experiments with four models and multiple agent harnesses, agents violated authorization in 19.2% of trajectories, with rates ranging from 10.5% to 26.6% across configurations, and increased reasoning effort did not substantially reduce these violations.
arXiv:2610.24993v2 Announce Type: replace Abstract: The throughput gains of speculative decoding depend on the trade-off between acceptance length and execution latency. We ask when additional accept...
arXiv:2610.27904v1 Announce Type: new Abstract: Staged post-training of multimodal foundation models chains vision-language alignment, reasoning distillation, and domain tuning, yet how each phase tr...
arXiv:2610.27921v1 Announce Type: new Abstract: A single observation can admit several plausible futures, making uncertainty an essential part of world modelling. However, quantum generative models r...
The paper introduces a two‑stage framework for reflective prompt optimization that separates candidate generation from validation. In the construction stage, multiple parents are sampled and their training responses reused across parallel reflection branches, with each revision screened on a small minibatch to allow locally improved candidates to propagate without full validation. In the exploration stage, a fixed full‑validation budget is used to select the best prompt from a frozen candidate pool, enabling independent control of pool size and validation effort. Experiments on five benchmarks with Qwen3.5‑9B and Llama‑3.1‑8B‑Instruct show that retaining 100 candidates and performing 20 full validations can improve performance by up to 11.22 percentage points over the initial solver, and learned selection consistently outperforms the initial solver across all comparisons.
arXiv:2610.28074v1 Announce Type: new Abstract: As a prototypical latent world model, Joint-Embedding Predictive Architectures (JEPAs) provide strong semantic representations. However, dense patch re...
arXiv:2610.28402v1 Announce Type: new Abstract: As streaming video generation advances, enabling real-time interaction with generated video is becoming increasingly important. We introduce StreamEval...
arXiv:2610.27606v1 Announce Type: cross Abstract: Offline reinforcement learning is fundamentally limited by the coverage of the available data, while collecting additional trajectories through onlin...
The paper presents a bounded, online step-level repair module for Lean 4 that uses compiler error messages to fix rejected tactics instead of discarding them. By sampling up to $K$ candidate tactics and allowing at most $R$ repairs per step, the method improves solve rates over a sampling-only baseline, achieving up to a 32% relative gain and solving 82 additional theorems on a 1,990-theorem subset of the LeanDojo benchmark. The approach trades extra runtime for reachable proofs, with most benefit coming from early repair attempts and shallow tactic failures, and it successfully replicates on the Lean Workbook benchmark, nearly doubling its solve rate without regressions.
arXiv:2610.27298v1 Announce Type: cross Abstract: Class-Incremental Learning (CIL) aims to learn new classes continuously without forgetting the previously learned ones. Recent works in CIL have sign...
The paper investigates how language models fail to detect harmful intent after polarity-altering transformations, a vulnerability called polarity blindness. It examines grammatical negation, antonym replacement, and synonym replacement as controls, and tests the impact of adversarial wrappers. Across 260 prompts, 10 LLMs, and three languages, polarity blindness is widespread, with the most vulnerable model scoring an average harm level of 4.12 on a 0‑5 scale.
arXiv:2610.28411v1 Announce Type: cross Abstract: Deep learning models have become essential tools in medical imaging, yet training them end-to-end for diverse tasks remains challenging due to limite...
arXiv:2610.28607v1 Announce Type: cross Abstract: Mathematical reasoning remains underexplored for low-resource languages, including Dialectal Arabic. To bridge this gap, we introduce the DAR-MGSM (D...
arXiv:2610.28562v1 Announce Type: new Abstract: Many scientific phenomena are best understood as outcomes of underlying optimization problems, yet automated scientific discovery has largely focused o...
arXiv:2610.28584v1 Announce Type: new Abstract: Autoregressive models (ARMs) face a "Sequentiality Bottleneck" in reasoning, while diffusion language models (DLMs) offer parallelism but lack semantic...
arXiv:2610.26992v1 Announce Type: cross Abstract: Split Federated Learning (SFL) has emerged as a promising paradigm for training large-scale deep models on resource-constrained edge devices. Althoug...
WaveCommitments is an open‑source framework that enables private and verifiable marketplaces for autonomous speech agents. It combines confidential asset valuation using Data Shapley within a trusted execution environment with multivariate polynomial commitments, allowing selective openings and tamper‑evident provenance. Experiments on three datasets and state‑of‑the‑art encoders, including Whisper, show up to 20× faster benchmarks and 39× smaller proofs, while also revealing how speaker value varies across tasks and encoders.
HierDTA is a hierarchical structure‑aware framework for drug–target affinity prediction that represents drugs as atom–motif heterogeneous graphs and proteins as residue interaction graphs. It integrates atom and residue representations via cross‑attention, supplemented by molecule fingerprints and pretrained ESM‑2 embeddings, achieving superior performance on the Davis and KIBA datasets and strong generalization in cold‑start scenarios. Ablation studies and interpretability analyses confirm the contributions of its hierarchical augmentation and interaction components, and case studies demonstrate early enrichment of candidate bioactive compounds.