AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Learning Chain of Counterfactual Thought for Bias-Robust Vision-Language Reasoning

Yifeng Zhang, Ming Jiang, Qi Zhao

ECCV 2024 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 2 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 the TLM reasoning-process predictor to CoCT (ablation)CoCT on LLaVA-1.5 · Bingoaccuracy: 18.8 -> 29.8Improved
add bias-robust loss L_br to TLM training (ablation)CoCT on LLaVA-1.5 · CoBRa counterfactual splitaccuracy: 34.2 -> 38.3Improved
include both original and counterfactual in-context examples instead of one type (ablation)CoCT on LLaVA-1.5 · CoBRa counterfactual splitaccuracy: 33.7 -> 38.3Improved
select in-context examples by reasoning similarity rather than knowledge similarity (ablation)CoCT on LLaVA-1.5 · CoBRa counterfactual splitaccuracy: 31.2 -> 32.1Improved
knowledge-similarity selection instead of reasoning similarity (ablation)CoCT on LLaVA-1.5 · ScienceQA-IMGaccuracy: 70.8 vs 70.1 (random)Mixed
add reasoning processes to the prompt (ablation)CoCT on LLaVA-1.5 · CoBRa gap (delta)delta (lower better): 44.2 -> 38.1Improved
add knowledge graphs to the prompt (ablation)CoCT on LLaVA-1.5 · CoBRa counterfactual splitaccuracy: 29.9 -> 38.3Improved
add in-context examples versus zero-shot (ablation)CoCT on LLaVA-1.5 · HallusionBenchaccuracy: 29.4 -> 30.7 (random)Improved
apply CoCT (CoBRa examples) over plain LVLMLLaVA-1.5 · Bingoaccuracy: 12.5 -> 29.8Improved
use CoCT over DDCoT on ScienceQALLaVA-1.5 and GPT-4V · ScienceQA-IMGaccuracy: 74.6 -> 71.3 (LLaVA); 80.2 -> 75.8 (GPT-4V)Hurt
use CoBRa instead of GQA as example sourceCoCT · GQA-OOD / VQA-CP (LLaVA-1.5)accuracy: VQA-CP 56.2 -> 56.0Mixed

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