Share your thoughts, 1 month free Claude Pro on usSee more

Instruction Following on ComplexBench

85.22Overall Score

ImpRIF-32B

Updated 1mo ago

Evaluation Results

Method	Links
ImpRIF-32B 2026.02		85.22
ImpRIF-32B 2026.02		84.1
Qwen3-235B-A22B 2026.02		83.46
ImpRIF-32B 2026.02		83.39
ImpRIF-8B 2026.02		83.29
Qwen3-14B 2026.02		82.16
Qwen3-32B 2026.02		81.99
ImpRIF-8B 2026.02		81.56
Qwen2.5-72B 2026.02		81.48
Qwen3-8B 2026.02		81.37
ImpRIF-8B 2026.02		81.22
ImpRIF-4B 2026.02		80.91
ImpRIF-4B 2026.02		79.37
ImpRIF-4B 2026.02		77.57
Qwen3-4B 2026.02		77.04
Ministral3-14B 2026.02		76.8
Llama3.1-70B 2026.02		73.76
MuSC 2025.02		0.7
MuSC 2025.02		0.6939
Self-Reward w/ BSM 2025.02		0.6743
ISHEEP 2025.02		0.6732
Qwen2-7B-Instruct 2025.02		0.6724
ISHEEP 2025.02		0.6713
Self-Reward w/ BSM 2025.02		0.6702
Self-Reward 2025.02		0.6698
Self-Reward 2025.02		0.6645
MuSC 2025.02		0.6598
SFT 2025.02		0.6589
MuSC 2025.02		0.6473
Self-Correct 2025.02		0.6441
Self-Correct 2025.02		0.6432
Self-Reward w/ BSM 2025.02		0.643
Self-Reward w/ BSM 2025.02		0.6413
Self-Reward w/ GPT-4 2025.02		0.6405
Self-Reward w/ GPT-4 2025.02		0.6352
ISHEEP 2025.02		0.6292
ISHEEP 2025.02		0.6267
Self-Reward 2025.02		0.6245
Self-Reward 2025.02		0.6203
LLaMA-3-8B-Instruct 2025.02		0.6149
Self-Correct 2025.02		0.6079
Self-Correct 2025.02		0.5591
SFT 2025.02		0.5393