AI Paper Atlas

Research map / Video / Video-language understanding

Adapting MLLMs for Nuanced Video Retrieval

Piyush Nitin Bagad, Andrew Zisserman

ECCV 2026 · Video › Video-language understanding · approach: Text-video retrieval · 1 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
fine-tune MLLM on 20K text-only triplets with nuance hard negatives (TARA)Tarsier 2 · CiA SSv2 (Chiral)mAP: 77.7 -> 88.9Improved
apply TARA to autoregressive base MLLMsQwen2VL / InternVL3 · CiA SSv2 (Chiral)mAP: 56.8 -> 73.1 (InternVL3); 60.2 -> 70.1 (Qwen2VL)Improved
apply TARA to already contrastively trained Qwen3VL-EmbeddingQwen3VL-Embedding · CiA SSv2 (Chiral)mAP: 72.0 -> 73.0Improved
TARA on Qwen3VL-Embedding for EPIC staticQwen3VL-Embedding · CiA EPIC (Static)mAP: 28.6 -> 25.1Hurt
text-only TARA fine-tuning vs multimodal fine-tuning on same Qwen2VL baseQwen2VL-7B · Nuanced retrieval averageavg score: 51.9 -> 56.8Improved
TARA zero-shot on negation queriesTarsier 2 · NegBench COCOR-Neg@5: 21.5 -> 73.6Improved
TARA zero-shot on composed video retrievalTarsier 2 · WebVid-CoVRR@1: 59.8 -> 66.3Improved
TARA on standard video benchmarksTarsier 2 · MMEB-V2 video classificationAvg.: 27.4 -> 66.1Improved
TARA vs Qwen3-VL-Embedding on standard benchmarksTarsier 2 + TARA · MMEB-V2 video retrievalAvg.: 45.9 vs 58.9Hurt
concatenate TARA and Qwen3VL-Embedding embeddingsQwen3VL-Embedding · MMEB-V2 video retrievalAvg.: 58.9 -> 59.5Improved
increase training data size from 5K to 200K text triplets (ablation)TARA · CiA SSv2 (Chiral)mAPNo clear effect
use EOL prompt alone to close modality gap (no fine-tuning) (ablation)Qwen2VL · MSRVTT video-textmodality gapNo clear effect

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