Verified75% confidenceFactExact time
The off-policy distillation experiment uses Qwen3 0.6B as the student model and Qwen3 4B as the teacher model.
3
Sources
75%
Confidence
Medium-term (~90 days)
Relevance
9/12/2026
First Seen
Valid until: 12/11/2026
Sources
Related Entities
Related Claims
UnverifiedQwen 3 is an open-source large language model series released by Alibaba's Tongyi Lab, featuring hybrid reasoning modes and reinforcement learning post-training capabilities.75% similarUnverifiedThe Qwen series of models was developed by Alibaba's Tongyi Lab.71% similarUnverifiedThe Qwen team has open-sourced QwenGyre, an elastic reinforcement learning framework for ultra-long-horizon agent training (XLong agent training).66% similarUnverifiedResearchers tested two technical approaches on the Qwen3-14B model: directly training the model to learn when to answer, and training an independent answer selector64% similarUnverifiedThe researchers eliminated the dependency on SenseNova at inference time by training a distilled student network with the architecture 5120→512→512→5120 (RMS normalization + SiLU activation) to reconstruct the projected teacher representations from H3's L49 conditions.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/911349API
curl https://kongchang.com/api/v1/knowledge/claims/911349MCP
get_claim(id=911349)