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

Synthetic Long-context Reasoning on LongReason

84.89Accuracy

Qwen3-30B-A3B-Thinking-2507 + Data Recipe for Long-Context RL

Updated 1mo ago

Evaluation Results

Method	Links
Qwen3-30B-A3B-Thinking-2507 + Data Recipe for Long-Context RL 2026.06		84.89
Qwen3-30B-A3B-Thinking-2507 + KeyChain-15K 2026.06		84.76
Qwen3-4B-Thinking-2507 + Data Recipe for Long-Context RL 2026.06		82.12
Qwen3-30B-A3B-Thinking-2507 2026.06		81.61
Qwen3-30B-A3B-Thinking-2507 + DocQA-RL-1.6K 2026.06		80.86
Qwen3-4B-Thinking-2507 + KeyChain-15K 2026.06		76.2
Qwen3-8B-128K + Data Recipe for Long-Context RL 2026.06		73.93
Qwen3-8B-128K + KeyChain-15K 2026.06		72.04
Qwen3-4B-Thinking-2507 + DocQA-RL-1.6K 2026.06		71.28
Qwen3-4B-Thinking-2507 2026.06		69.5
Qwen3-8B-128K 2026.06		68.26
Qwen3-8B-128K + DocQA-RL-1.6K 2026.06		67.88