General
ComplexBench
ComplexBench: complex instruction following with multiple composed constraints (NeurIPS 2024). Each instruction decomposes into binary scoring questions; this build ingests the deterministic, rule-verified subset of those questions, grading each released model generation with the benchmark's own rule evaluator.
252items
15subjects
MITlicense
generaldomain
textmodality
item-level responses released
Saturation status: Unknown
Response matrix
Fit to width. Hover for subject & item; click a cell for details.
Correct (1)Incorrect (0)Unobserved
Scale: 1 = correct · 0 = incorrect
Subjects
- 1gpt3.5_turbo_11060.7063
- 2erniebot_40.696
- 3gpt4_11060.6865
- 4glm40.6825
- 5claude3_opus0.6786
- 6qwen1.5_72b0.619
- 7qwen1.5_14b0.5952
- 8llama3_70b0.5833
- 9qwen1.5_7b0.5833
- 10internlm2_20b0.5833
- 11baichuan2_13b0.5635
- 12chatglm3_6b0.5635
- 13internlm2_7b0.5595
- 14llama3_8b0.5119
- 15mistral_7b0.4603