InftyThink: Breaking the Length Limits of Long-Context Reasoning in Large Language Models
Paper • 2503.06692 • Published • 2
InftyThink+
InftyThink+: Effective and Efficient Infinite-Horizon Reasoning via Reinforcement Learning
-
-
InftyThink+: Effective and Efficient Infinite-Horizon Reasoning via Reinforcement Learning
Paper • 2602.06960 • Published • 14
yanyc/InftyThink-1.5B
2B • Updated • 10 • 1Note SFT model trained with InftyThink-style OpenThoughts-114k from DeepSeek-R1-Distill-Qwen-1.5B.
yanyc/InftyThink-Plus-T-1.5B
2B • Updated • 11Note RL model trained with DeepScaleR from yanyc/InftyThink-1.5B using Task reward.
yanyc/InftyThink-Plus-TE-1.5B
2B • Updated • 12Note RL model trained with DeepScaleR from yanyc/InftyThink-1.5B using Task reward and Efficiency reward.
yanyc/InftyThink-4B
4B • Updated • 7Note SFT model trained with InftyThink-style OpenThoughts-114k from Qwen3-4B-Base.
yanyc/InftyThink-Plus-T-4B
4B • Updated • 7Note RL model trained with DeepScaleR from yanyc/InftyThink-4B using Task reward.
yanyc/InftyThink-Plus-TE-4B
4B • Updated • 9Note RL model trained with DeepScaleR from yanyc/InftyThink-4B using Task reward and Efficiency reward.
yanyc/OpenThoughts-114k-InftyThink
Viewer • Updated • 178k • 23Note InftyThink-style cold start data.