Research map / Large language models / Retrieval and knowledge
VaccineRAG: Boosting Multimodal Large Language Models’ Immunity to Harmful RAG Samples
Qixin Sun, Ziqin Wang, Hengyuan Zhao, Yilin Li, Kaiyou Song, Si Liu, Xiaolin Hu, Qingpei Guo, Linjiang Huang
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 helpfulness reward (ablation) | Partial GRPO (Qwen2-VL-7B) · WebQA polluted generation (VaccineRAG) | MA5: 52.71 -> 56.27 | Improved |
| add conclusion reward (ablation) | Partial GRPO (Qwen2-VL-7B) · WebQA polluted generation (VaccineRAG) | MA5: 53.90 -> 56.27 | Improved |
| add helpfulness reward (ablation) | Partial GRPO (Qwen2.5-VL-7B) · WebQA polluted generation (VaccineRAG) | MA5: 58.44 -> 64.53 | Improved |
| add conclusion reward (ablation) | Partial GRPO (Qwen2.5-VL-7B) · WebQA polluted generation (VaccineRAG) | MA5: 60.29 -> 64.53 | Improved |
| add helpfulness reward (ablation) | Partial GRPO (InternVL3-8B) · WebQA polluted generation (VaccineRAG) | ADR5: 31.18 -> 11.76 | Improved |
| add conclusion reward (ablation) | Partial GRPO (InternVL3-8B) · WebQA polluted generation (VaccineRAG) | MA5: 58.68 -> 62.28 | Improved |
| add format reward (ablation) | Partial GRPO · WebQA polluted generation (VaccineRAG) | training stability | Improved |
| replace vanilla GRPO with Partial GRPO | Qwen2-VL-7B · WebQA polluted generation (VaccineRAG) | MA5: 54.50 -> 56.27 | Improved |
| replace vanilla GRPO with Partial GRPO | Qwen2.5-VL-7B · WebQA polluted generation (VaccineRAG) | MA5: 59.83 -> 64.53 | Improved |
| replace vanilla GRPO with Partial GRPO | Qwen2.5-VL-7B · WebQA polluted generation (VaccineRAG) | ADR5: 0.21 -> 10.43 | Hurt |
| replace vanilla GRPO with Partial GRPO | InternVL3-8B · WebQA polluted generation (VaccineRAG) | MA5: 60.61 -> 62.28 | Improved |
| add CoT to SURf | Qwen2-VL-7B · WebQA polluted generation (VaccineRAG) | MA5: 45.28 -> 47.58 | Improved |
Closest papers by meaning
- RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs (NeurIPS 2024 · similarity 0.91)
- Long-Context LLMs Meet RAG: Overcoming Challenges for Long Inputs in RAG (ICLR 2025 · similarity 0.90)
- Benchmarking Large Language Models in Retrieval-Augmented Generation (AAAI 2024 · similarity 0.90)
- One Token Can Help! Learning Scalable and Pluggable Virtual Tokens for Retrieval-Augmented Large Language Models (AAAI 2025 · similarity 0.90)
- Retrieval meets Long Context Large Language Models (ICLR 2024 · similarity 0.89)
- Inference Scaling for Long-Context Retrieval Augmented Generation (ICLR 2025 · 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.
