Research map / Vision-language and multimodal / Multimodal LLMs
GenRecal: Generation after Recalibration from Large to Small Vision-Language Models
Byung-Kwan Lee, Ryo Hachiuma, Yong Man Ro, Yu-Chiang Frank Wang, Yueh-Hua Wu
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 |
|---|---|---|---|
| distill with GenRecal from InternVL2.5-78B | InternVL2.5-8B · MMMU | accuracy: 56.0 -> 68.1 | Improved |
| replace LLaVA-KD with GenRecal under same token type | Qwen2-VL-7B · MMMU | accuracy: 58.2 -> 64.2 | Improved |
| add first-stage regularization (ablation) | GenRecal (InternVL2.5-8B) · MMMU | accuracy: 58.9 -> 68.1 | Improved |
| add first-stage regularization (ablation) | GenRecal (InternVL2.5-1B) · MMMU | accuracy: 41.1 -> 45.6 | Improved |
| add autoregressive loss to KL loss (ablation) | GenRecal (InternVL2.5-8B) · Avg over 11 benchmarks | average accuracy: 59.8 -> 77.1 | Improved |
| add new positional embedding (NPE) (ablation) | GenRecal (Qwen2-VL-7B) · Avg of MMB/MathVista/MM-Vet/MMMU | average accuracy: 66.8 -> 73.3 | Improved |
| increase Recalibrator depth from 2 to 20 (ablation) | GenRecal · Avg of MMB/MathVista/MM-Vet/MMMU | average accuracy: 76.4 -> 76.7 | No clear effect |
| reduce Recalibrator depth from 2 to 1 (ablation) | GenRecal · Avg of MMB/MathVista/MM-Vet/MMMU | average accuracy: 76.4 -> 74.9 | Hurt |
| increase training data from 5M to 9M (ablation) | GenRecal · Avg of MMB/MathVista/MM-Vet/MMMU | average accuracy: 76.0 -> 76.4 | No clear effect |
| increase training data from 1M to 9M (ablation) | GenRecal · Avg of MMB/MathVista/MM-Vet/MMMU | average accuracy: 70.8 -> 76.4 | Improved |
| use GenRecal instead of cross-tokenizer MOT (ablation) | InternVL2.5-8B · Avg of MMB/MathVista/MM-Vet/MMMU | average accuracy: 70.8 -> 76.4 | Improved |
| use a stronger teacher VLM (ablation) | GenRecal · MMMU | accuracy: 60.3 (NVLM-72B) -> 68.1 (InternVL2.5-78B) | Improved |
Closest papers by meaning
- What matters when building vision-language models? (NeurIPS 2024 · similarity 0.90)
- Are We on the Right Way for Evaluating Large Vision-Language Models? (NeurIPS 2024 · similarity 0.90)
- VHELM: A Holistic Evaluation of Vision Language Models (NeurIPS 2024 · similarity 0.90)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.89)
- LLMC+: Benchmarking Vision-Language Model Compression with a plug-and-play Toolkit (AAAI 2026 · similarity 0.89)
- Mordal: Automated Pretrained Model Selection for Vision Language Models (ICLR 2026 · similarity 0.89)
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.
