List your company
ResearchFormerly WizWand

Ad-Injected Response Generation on MT-Human

88.8AccuracyGI-R
64.46470.78277.183.418Sep 17, 2025
Updated 5mo ago

Evaluation results

MethodLinks
GI-RJudgement LLM=claude-3-5-haiku, Base LLM=doubao-1-5-lite2025.0988.849.882.875.632.462.465.3
GIR-RJudgement LLM=claude-3-5-haiku, Base LLM=doubao-1-5-lite2025.0988.253.478.674.443.266.667.4
GIR-PJudgement LLM=claude-3-5-haiku, Base LLM=doubao-1-5-lite2025.0987.65476.871.429.465.464.1
GIR-RBase LLM=doubao-1.5-lite-32k, Evaluator LLM=gpt-4.1-mini, Embedding model=text-embedding-3-small, Retrieval strategy=response-based, Final rewriting=included2025.0985.852.874.473.282.879.274.7
GI-RJudgement LLM=kimi-k22025.0984.636.671.446.225.252.852.8
GI-RBase LLM=doubao-1.5-lite-32k, Evaluator LLM=gpt-4.1-mini, Embedding model=text-embedding-3-small, Retrieval strategy=response-based, Final rewriting=skipped2025.0983.439.679.868.474.472.669.7
GIR-RJudgement LLM=kimi-k22025.0983.446.27261.246.25761
GIR-PBase LLM=doubao-1.5-lite-32k, Evaluator LLM=gpt-4.1-mini, Embedding model=text-embedding-3-small, Retrieval strategy=prompt-based, Final rewriting=included2025.0982.853.474.470.281.68173.9
GI-RJudgement LLM=qwen-max2025.0982.241.47554.649.243.857.7
GIR-PJudgement LLM=kimi-k22025.0981.646.870.854.63652.857.1
GIR-RJudgement LLM=qwen-max2025.098153.464.252.863.655.261.7
GIR-PJudgement LLM=qwen-max2025.0976.252.264.256.46353.460.9
Ad-ChatJudgement LLM=claude-3-5-haiku, Base LLM=doubao-1-5-lite2025.0976.243.257.652.831.853.452.5
Ad-ChatBase LLM=doubao-1.5-lite-32k, Evaluator LLM=gpt-4.1-mini, Configuration=Original implementation2025.0972.65168.461.872.664.265.1
Ad-ChatJudgement LLM=qwen-max2025.0971.451.662.453.455.846.856.9
Ad-ChatJudgement LLM=kimi-k22025.0965.443.256.449.825.840.846.9