AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

AURELIA: Test-time Reasoning Distillation in Audio-Visual LLMs

Sanjoy Chowdhury, Hanan Gani, Nishit Anand, Sayan Nag, Ruohan Gao, Mohamed Elhoseiny, Salman Khan, Dinesh Manocha

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 9 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add AURELIA reasoning distillationVITA · AVReasonBench AV-GeoIQaccuracy: 26.7 -> 39.0Improved
add AURELIA reasoning distillationVITA · AVReasonBench AV-Captioning (CIDEr)CIDEr: 73.8 -> 78.8Improved
add AURELIA reasoning distillationVITA · AVReasonBench AV-Memeaccuracy: 44.0 -> 50.0Improved
add AURELIA reasoning distillationX-InstructBLIP · AV-GeoIQaccuracy: 14.5 -> 29.0Improved
add AURELIA reasoning distillationVITA · Video-MME overallaccuracy: 80.7 -> 83.73Improved
add AURELIA reasoning distillationVITA · Video-MME long videosaccuracy: 73.3 -> 76.98Improved
add AURELIA reasoning distillationVITA · AV-CaptioningBLEU@4: 12.9 -> 14.5Improved
add AURELIA reasoning distillationPandaGPT · AV-GeoIQaccuracy: 12.5 -> 25.0Improved
use GPT-4o as summarizer instead of Gemini for all agents (ablation)AURELIA (VideoLLaMA-2) · AV-GeoIQaccuracy: 36.5 -> 38.0Improved
use GPT-4o as summarizer instead of Gemini for all agents (ablation)AURELIA (VideoLLaMA-2) · DM-Matchaccuracy: 33.0 -> 34.5Improved
increase feedback iterations from 1 to 5 (ablation)AURELIA (VideoLLaMA-2) · AV-GeoIQaccuracy: 27.5 -> 38.0Improved
increase iterations from 1 to 5 (ablation)AURELIA (VideoLLaMA-2) · AV-Compaccuracy: 27.3 -> 31.6Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy