Alignment, interpretability, red-teaming, bias and privacy: the research on what these systems do when they misbehave.
arXiv:2610.17880v1 Announce Type: cross Abstract: Persona vectors are linear directions in the residual-stream activations of LLMs that encode behavioral traits and enable fine-grained control over m...
arXiv:2610.17706v1 Announce Type: cross Abstract: Large vision-language models often exhibit object hallucination, generating claims that are not supported by the input image. To mitigate this issue,...
arXiv:2610.16761v1 Announce Type: new Abstract: LLM-based information systems need safety guardrails. Most guardrails classify the prompt alone. If the prompt looks harmful, the system rejects it bef...
arXiv:2610.17197v1 Announce Type: cross Abstract: In the age of "Software 3.0", when a complete program can be rewritten in minutes by an agent, the traditional approach of relying on code as the sol...
arXiv:2610.17113v1 Announce Type: cross Abstract: Information extraction models often perform well on the datasets they are trained on, but transferring to new domains and annotation guidelines remai...
arXiv:2610.14697v1 Announce Type: cross Abstract: Speech-conditioned generation maps speech signals to target audio outputs. Prior end-to-end speech-conditioned generative models achieve high fidelit...
arXiv:2610.15056v1 Announce Type: cross Abstract: Large language models can post-edit ASR by using linguistic context to compare N-best hypotheses and rewrite the transcript when needed, often improv...
arXiv:2610.16694v1 Announce Type: cross Abstract: Adversarial detection on evolving attributed graphs faces two challenges: fraud labels arrive late or not at all, and natural distributional drift er...
arXiv:2610.16946v1 Announce Type: cross Abstract: Descriptions of how people use AI are becoming training data for the next generation of models. We ask whether what a model reads about its users cha...
arXiv:2610.19543v1 Announce Type: cross Abstract: In medical annotation and crowdsourcing, multiple annotators assign labels to each example, but these labels may be incorrect. Treating such noisy la...
arXiv:2610.19035v1 Announce Type: new Abstract: Pluralistic alignment needs to accommodate diverse preferences under limited human feedback. Using AI simulators to label preference comparisons reduce...
arXiv:2610.19133v1 Announce Type: new Abstract: Indirect prompt injection uses untrusted retrieved content, tool outputs, or agent messages to override an LLM's intended behavior. We show that defens...
arXiv:2610.19191v1 Announce Type: new Abstract: Wildfire decisions are made in hours, yet most learned spread models predict daily on targets including burned area. We propose FireGraphNet which form...
arXiv:2601.02978v3 Announce Type: replace-cross Abstract: Recent work in Mechanistic Interpretability (MI) has enabled the identification and intervention of internal features in Large Language Model...
arXiv:2601.12243v3 Announce Type: replace-cross Abstract: Video summarization transforms long videos into concise representations that are easier to document and analyze, especially in high-stakes do...
arXiv:2601.17673v2 Announce Type: replace-cross Abstract: Unified remote sensing multimodal models exhibit a pronounced spatial reversal curse: Although they can accurately recognize and describe obj...
arXiv:2601.20055v3 Announce Type: replace-cross Abstract: Despite the syntactic fluency of Large Language Models (LLMs), ensuring their logical correctness in high-stakes domains remains a fundamenta...
arXiv:2610.19892v1 Announce Type: new Abstract: Vision-language models suffer significant accuracy drops under common corruptions. While test-time adaptation (TTA) successfully mitigates these shifts...
arXiv:2610.14701v1 Announce Type: cross Abstract: Model merging has become widely adopted for enhancing the task-specific performance of Large Language Models (LLMs) without additional training. Whil...
arXiv:2610.14802v1 Announce Type: cross Abstract: End-to-end driving policies can achieve strong task performance while offering limited control over their driving style. Driving style describes how...