List your company

Mathematical Reasoning on AMOBench (Pass@16)

61.5Pass@16TRD
8.25222.07635.949.724Jun 7, 2026

Evaluation results

MethodLinks
TRDBackbone=Qwen3-8B (w/ thinking), Trajectory=yr2026.0661.5
Forward KLBackbone=Qwen3-8B (w/ thinking), Trajectory=yo2026.0651.3
Reverse KLBackbone=Qwen3-8B (w/ thinking), Trajectory=yo2026.0651.3
Forward KL w/ ClipBackbone=Qwen3-8B (w/ thinking), Trajectory=yo2026.0643.6
BaseBackbone=Qwen3-8B (w/ thinking), Trajectory=–2026.0641
Reverse KL w/ Top-KBackbone=Qwen3-8B (w/ thinking), Trajectory=yo2026.0638.5
TRDModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yr, Teacher Model=Qwen3-8B2026.0635.9
TRDBackbone=Qwen3-4B-Instruct-2507, Trajectory=yr2026.0633.8
Forward KLBackbone=Qwen3-4B-Instruct-2507, Trajectory=yo2026.0633.3
Forward KLModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.0633.3
Forward KL w/ ClipBackbone=Qwen3-4B-Instruct-2507, Trajectory=yo2026.0628.2
Reverse KLBackbone=Qwen3-4B-Instruct-2507, Trajectory=yo2026.0628.2
Reverse KL w/ Top-KModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.0628.2
Reverse KL w/ Top-KBackbone=Qwen3-4B-Instruct-2507, Trajectory=yo2026.0625.6
Forward KL w/ ClipModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.0625.6
BaseBackbone=Qwen3-4B-Instruct-2507, Trajectory=–2026.0623.1
BaseModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=-, Teacher Model=Qwen3-8B2026.0623.1
TRDModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yr, Teacher Model=Qwen3-8B2026.0617.9
Reverse KLModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.0617.9
Reverse KL w/ Top-KModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.0615.4
BaseModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=-, Teacher Model=Qwen3-8B2026.0612.8
Forward KLModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.0612.8
Reverse KLModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.0612.8
Forward KL w/ ClipModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.0610.3