List your company

Multi-round coreference resolution on MRCR

81.62-Needle ScoreDeepSeek-V3.2-Thinking
45.324854.742464.1673.5776May 21, 2026

Evaluation results

MethodLinks
DeepSeek-V3.2-Thinking2026.0581.660.3271.01
Qwen3-30B-A3B-Thinking + ACC2026.0576.959.5768.28
Qwen3-235B-A22B-Thinking2026.0574.9859.9667.51
GPT-5.1-Thinking2026.0573.7547.9360.91
GLM-4.6-Thinking2026.0573.6355.3364.53
Kimi-K2-Thinking2026.0568.0147.9658.05
Qwen3-30B-A3B-Thinking2026.0561.8438.4150.19
GPT-OSS-120B2026.0546.7229.1638