·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
SciReC: Diagnostic Evaluation of Multimodal, Multi-Turn Relational Reasoning with Adaptive Interaction1h◆UIC-AIHealth4All at ArchEHR-QA 2026: Answer-First Evidence Grounding for Clinical Question Answering1h◆Select, Don't Train: The Benefits of Modular Entity Disambiguation with LLM-Based Selection1h◆INSPIRE: An Internalize-Then-Improve Approach for Example-Driven Mathematical Reasoning1h◆How Do Linear Probes Emerge? A Circuit-Tracing Framework with Concept-Targeted Attribution1h◆Trajectory-Level Speculative Decoding for Diffusion Language Models1h◆Below the Noise Floor: Bimodal Seed Collapse and Distinct Failure Modes in Small-Model Knowledge Distillation1h◆Load-Bearing Context: The Question Damage Score for Evaluating Context Reliance in Linguistic Reasoning1h◆Informational Antilocality and the Locality Bias in LLMs1h◆EvoHarmBench: Breaking Content Moderation with Iterative Human-Like Evasion1h◆OpenStamp: A Watermark for Open-Source Language Models1h◆Lexically conditioned realization ambiguity in Korean predicate morphology1h◆QUORUM: QUality-Optimized Routing Using Multiple annotators1h◆Predicting Turn-Taking Outcomes in Multi-Party Conversation: Interpretable Modelling of Speech and Gaze Dynamics with Interpersonal Closeness1h◆When Linguistic and Internal Confidence Diverge in Large Language Models1h◆CultureConverse: A Multilingual Multi-turn Simulation Harness for Culturally Grounded Assistance in East and Southeast Asia1h◆A Unified Framework to Elicit Structured Feedback for Interpretable Multi-Trait Essay Scoring1h◆PACE: Publisher-Adaptive Content Extraction via Agentic Automation1h◆Fidelity Is Not Enough: Dispatch-Level Instrumentation for Agentic Datasheet Extraction1h◆Quantifying Affective Bias in Low-Resource Media: Large-Scale Emotion Profiling of Bengali Headlines1h◆SciReC: Diagnostic Evaluation of Multimodal, Multi-Turn Relational Reasoning with Adaptive Interaction1h◆UIC-AIHealth4All at ArchEHR-QA 2026: Answer-First Evidence Grounding for Clinical Question Answering1h◆Select, Don't Train: The Benefits of Modular Entity Disambiguation with LLM-Based Selection1h◆INSPIRE: An Internalize-Then-Improve Approach for Example-Driven Mathematical Reasoning1h◆How Do Linear Probes Emerge? A Circuit-Tracing Framework with Concept-Targeted Attribution1h◆Trajectory-Level Speculative Decoding for Diffusion Language Models1h◆Below the Noise Floor: Bimodal Seed Collapse and Distinct Failure Modes in Small-Model Knowledge Distillation1h◆Load-Bearing Context: The Question Damage Score for Evaluating Context Reliance in Linguistic Reasoning1h◆Informational Antilocality and the Locality Bias in LLMs1h◆EvoHarmBench: Breaking Content Moderation with Iterative Human-Like Evasion1h◆OpenStamp: A Watermark for Open-Source Language Models1h◆Lexically conditioned realization ambiguity in Korean predicate morphology1h◆QUORUM: QUality-Optimized Routing Using Multiple annotators1h◆Predicting Turn-Taking Outcomes in Multi-Party Conversation: Interpretable Modelling of Speech and Gaze Dynamics with Interpersonal Closeness1h◆When Linguistic and Internal Confidence Diverge in Large Language Models1h◆CultureConverse: A Multilingual Multi-turn Simulation Harness for Culturally Grounded Assistance in East and Southeast Asia1h◆A Unified Framework to Elicit Structured Feedback for Interpretable Multi-Trait Essay Scoring1h◆PACE: Publisher-Adaptive Content Extraction via Agentic Automation1h◆Fidelity Is Not Enough: Dispatch-Level Instrumentation for Agentic Datasheet Extraction1h◆Quantifying Affective Bias in Low-Resource Media: Large-Scale Emotion Profiling of Bengali Headlines1h◆
News/Compositional Failure in Audio-Visual LLMs: Late-Layer Prior Dominance Under Cross-modal Conflict
arxiv
PublishedAugust 31, 2026 at 4:00 AM

Compositional Failure in Audio-Visual LLMs: Late-Layer Prior Dominance Under Cross-modal Conflict

Source
arxiv.orgfull article ↗
Read on arxiv→
Publisher summary· verbatim

arXiv:2608.27785v1 Announce Type: new Abstract: We study audio-visual conflict as a compositional generalization test for AV-LLMs: the model must combine synchronized but semantically incompatible audio and video evidence and decide whether the pair matches. On VideoLLaMA 2-7B-AV, three alignment co

Stay posted· Newsletter

A 5-min weekly brief — top movers, price watch, story of the week.

// no spam · unsubscribe one-click · free forever

Discussion
Source
↗
arxiv
Read original ↗All from arxiv →

No replies yet. Be first.

Source
↗
arxiv
Read original ↗All from arxiv →

Related coverage

More from ARXIV
arxivSciReC: Diagnostic Evaluation of Multimodal, Multi-Turn Relational Reasoning with Adaptive Interaction1harxivUIC-AIHealth4All at ArchEHR-QA 2026: Answer-First Evidence Grounding for Clinical Question Answering1harxivSelect, Don't Train: The Benefits of Modular Entity Disambiguation with LLM-Based Selection1harxivINSPIRE: An Internalize-Then-Improve Approach for Example-Driven Mathematical Reasoning1h
The Bubble Brief
WEEKLY

Read AI insights every Tuesday — top movers, new releases, story of the week.

// no spam · unsubscribe one-click · free forever

Originally published on arxiv ↗
HomeModelsNews