Research map / Vision-language and multimodal / Visual question answering
ReasonEdit: Editing Vision--Language Models using Human Reasoning
Jiaxing Qiu, Kaihua Hou, Roxana Daneshjou, Ahmed Alaa, Thomas Hartvigsen
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add reasoning entries to the codebook instead of answer entries only (ablation) | ReasonEdit · Entry-type ablation, LLaVA-1.5-7B | R-Gen: 0.65 -> 0.87 | Improved |
| add reasoning entries to the codebook instead of answer entries only (ablation) | ReasonEdit · Entry-type ablation, LLaVA-1.5-7B | CoE-Gen: 0.72 -> 0.82 | Improved |
| replace ReasonEdit by best prior editor with reasoning (IKE-COT) on R-Gen | ReasonEdit · FVQA, LLaVA-1.5-7B | R-Gen: 0.81 -> 0.87 | Improved |
| use ReasonEdit instead of IKE-COT | ReasonEdit · FVQA, Qwen3-VL-8B | CoE-Gen: 0.77 -> 0.95 | Improved |
| add human reasoning supervision (COT) to FT | FT · FVQA, LLaVA-1.5-7B | R-Gen: 0.39 -> 0.61 | Improved |
| add human reasoning (COT) to GRACE | GRACE · FVQA, LLaVA-1.5-7B | CoE-Gen: 0.54 -> 0.54 | No clear effect |
| add human reasoning (COT) to MEND | MEND · FVQA, LLaVA-1.5-7B | Acc: 0.52 -> 0.45 | Hurt |
| inject noisy statements (50%) into reasoning (ablation) | ReasonEdit · Noisy reasoning, LLaVA-1.5-7B | R-Gen: 0.86 -> 0.82 | No clear effect |
| replace 50% of reasoning facts with noise (ablation) | ReasonEdit · Noisy reasoning, LLaVA-1.5-7B | R-Gen: 0.86 -> 0.77 | Hurt |
| replace 50% of reasoning facts with noise (ablation) | ReasonEdit · Noisy reasoning, LLaVA-1.5-7B | CoE-Gen: 0.98 -> 0.93 | Hurt |
| add dual vision-language embedding instead of a single vision layer key (ablation) | ReasonEdit · codebook keys | image generality | Improved |
| add dual embedding instead of single language layer key (ablation) | ReasonEdit · codebook keys | text generality | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- GRACE — GRACE: Graph-Based Contextual Debiasing for Fair Visual Question Answering (ECCV 2024) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ReasonEdit: Towards Reasoning-Enhanced Image Editing Models (CVPR 2026 · similarity 0.92)
- Reasoning to Edit: Hypothetical Instruction-Based Image Editing with Visual Reasoning (ICML 2026 · similarity 0.89)
- Attribution Analysis Meets Model Editing: Advancing Knowledge Correction in Vision Language Models with VisEdit (AAAI 2025 · similarity 0.89)
- Envisioning Beyond the Pixels: Benchmarking Reasoning-Informed Visual Editing (NeurIPS 2025 · similarity 0.88)
- Learning Action and Reasoning-Centric Image Editing from Videos and Simulation (NeurIPS 2024 · similarity 0.88)
- ComprehendEdit: A Comprehensive Dataset and Evaluation Framework for Multimodal Knowledge Editing (AAAI 2025 · similarity 0.88)
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.
