Alignment, interpretability, red-teaming, bias and privacy: the research on what these systems do when they misbehave.
arXiv:2603.06859v4 Announce Type: replace Abstract: Credit assignment for a team of LLM agents, what each message was worth, has mostly been treated as prediction: a learned critic, a trajectory-leve...
arXiv:2610.06083v2 Announce Type: replace Abstract: Most adversarial attacks on deep reinforcement learning (DRL) assume white-box access to the victim policy, which rarely holds in practice. This pa...
arXiv:2605.14605v3 Announce Type: replace-cross Abstract: Model providers increasingly release the weights of large language models. Although these models are safety-aligned before release, their saf...
arXiv:2605.19352v2 Announce Type: replace-cross Abstract: Understanding how humans and artificial intelligence systems predict and plan by interacting with their environment is a fundamental challeng...
arXiv:2610.09163v1 Announce Type: new Abstract: Task-oriented conversational agents remain fragile under real world conversation scenarios as they rarely follow a predictable script, especially when...
arXiv:2610.09486v1 Announce Type: cross Abstract: Spoken language identification (LID) aims to recognize the target language regardless of accent. In practice, however, LID models fine-tuned from sel...
arXiv:2610.09981v1 Announce Type: cross Abstract: LLM routers pick a cheap or expensive model per request by its content, and many gateways and some cloud platforms can log that choice with content l...
arXiv:2610.09200v1 Announce Type: new Abstract: We introduce StyleFields, a DeepSDF-based architecture for high-fidelity 3D reconstruction that enables controllable geometric style mixing: the coarse...
arXiv:2610.09450v1 Announce Type: new Abstract: Pixel-space diffusion models avoid the lossy VAE of latent models, which suggests an advantage on downstream tasks where fine-grained detail matters. W...
arXiv:2610.10090v1 Announce Type: new Abstract: Collaborative perception extends the sensing range of autonomous vehicles, but its performance degrades when shared features arrive stale or incomplete...
arXiv:2610.09027v1 Announce Type: cross Abstract: Modern language model systems operate autonomously over increasingly long contexts containing untrusted text and images. Can an adversarial input con...
arXiv:2606. 04033v2 Announce Type: replace Abstract: The validation of advanced nuclear reactor designs and fuel concepts will require the design of new critical experiments with high neutronic similarity to the target technology.
The paper presents a system that combines coded police crash records with officer narratives to estimate uncoded crash factors in Texas from 2017‑2025. It uses an in‑context tabular foundation model (Kumo Tabular) to process coded data, a calibrated System One model (Jev) to read sampled narratives, and human judgments to recalibrate probabilities, producing validated population estimates for factors such as hydroplaning, medical episodes, fatigue, animals, and phone use. The approach demonstrates that narrative analysis uncovers more injury crashes than coded fields and that a single additional round of human review can significantly tighten estimate precision. "whyItMatters":"The system provides safety offices with more accurate counts, a discordance map, a validated re‑read list, and a reading budget, enabling better resource allocation for road‑safety interventions."
CM-DPO introduces a Constraint-Margin Direct Preference Optimization method that replaces the binary preference signal of traditional DPO with a continuous margin based on a deterministic symbolic verifier, scaled by violation severity. Hard and soft constraints are separated via a lexicographic objective to prevent hard constraints from being traded off against preferences. Using procedurally generated constraint profiles and minimal-edit distillation within the SynPlan-R framework, an 8B model fine‑tuned with CM‑DPO achieves high pass and solve rates on several planning benchmarks, matching multi‑agent systems with lower latency and outperforming GPT‑4o on unseen tasks.
arXiv:2610. 10499v1 Announce Type: new Abstract: Online learning is an attractive framework in many domains because it permits well-defined learning even when data are dependent or chosen adversarially.
The paper examines the safety of small language models (SLMs) deployed on resource‑constrained, on‑device platforms, focusing on the LLaMA‑2‑7B‑Chat model. It shows that safety‑sensitive behavior is concentrated in a sparse subset of parameters, particularly the MLP down_proj layer, and that altering only 0.19% of these weights can significantly affect safety metrics while leaving overall accuracy largely unchanged. The study suggests that targeted fault analysis and selective integrity protection could improve safety for on‑device, agentic language models.
The paper investigates safe reinforcement learning by focusing on the episodic-cost tail rather than just the mean cost. It introduces the use of γαρα_{0.1} (the average cost of the worst 10% of episodes) to classify policies as tail-safe. The authors evaluate standard algorithms on Safety-Gymnasium tasks, identify tail-unsafe policies, and explore constraint families to control tail violations while maintaining return.
The paper introduces CanonQ, a unified quantization‑aware training framework that separates source canonicalization from task‑aware adaptation to enable extreme low‑bit LLM compression. By applying fixed rotations and energy normalization, CanonQ maps heterogeneous tensors to canonical coordinates, allowing frozen Gaussian‑reference codebooks to be reused across layers and models. Joint training then adapts the network to the coupled errors of weight, activation, and KV cache quantization, achieving significant performance gains—up to 14.28× lower perplexity and 57.9% higher zero‑shot accuracy on LLaMA models, and substantial improvements on instruction‑tuned MobileLLM‑Pro‑1B. "whyItMatters":"The framework demonstrates that carefully designed canonicalization and joint training can dramatically improve the efficiency and accuracy of low‑bit quantized large language models, offering a practical path to deploy powerful LLMs with reduced computational and memory footprints."
arXiv:2610.09737v1 Announce Type: cross Abstract: When does distribution alignment help a frozen foundation-model embedding generalize across acoustic domains? For cross-domain mosquito-species class...
arXiv:2610.09863v1 Announce Type: cross Abstract: Standard information retrieval metrics, such as mean Average Precision (mAP), assess performance one query at a time, based on how the similarities b...