Poisoning Attack on RAG on HotpotQA (test)
100PoisonedRAG Success Rateqwen/qwen2.5-7b-instruct
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| qwen/qwen2.5-7b-instructEvaluation protocol=strict incorrect-answer matching2026.03 | 100 | 93 | 96 | 1 | 4 | 100 | 0 | |
| qwen/qwen2-7b-instructEvaluation protocol=strict incorrect-answer matching2026.03 | 100 | 96 | 89 | 22 | 6 | 100 | 0 | |
| ibm/granite-3.3-8b-instructEvaluation protocol=strict incorrect-answer matching2026.03 | 100 | 99 | 98 | 10 | 7 | 100 | 0 | |
| meta/llama-4-maverick-17b-128e-instructEvaluation protocol=strict incorrect-answer matching2026.03 | 100 | 95 | 100 | 16 | 5 | 100 | 0 | |
| meta/llama-3.1-8b-instructEvaluation protocol=strict incorrect-answer matching2026.03 | 98 | 92 | 92 | 1 | 8 | 98 | 0 | |
| meta/llama-3.3-70b-instructEvaluation protocol=strict incorrect-answer matching2026.03 | 95 | 94 | 90 | 17 | 5 | 100 | 5 | |
| openai/gpt-oss-120bEvaluation protocol=strict incorrect-answer matching2026.03 | 93 | 91 | 94 | 5 | 4 | 100 | 7 | |
| openai/gpt-oss-20bEvaluation protocol=strict incorrect-answer matching2026.03 | 90 | 94 | 85 | 3 | 6 | 97 | 7 |