PostDyn: checkpoints of intermediate training steps
Collection
Off- and on-policy SFT and RL for the Qwen and Olmo series, for studying how post-training changes the weights (Ongoing) • 2 items • Updated • 1
Xudong Zhu
XudongZhu
AI & ML interests
Mechanistic interpretability, Representation Learning
Recent Activity
upvoted a collection about 17 hours ago
PostDyn: checkpoints of intermediate training steps liked a model about 17 hours ago
XudongZhu/Qwen3-1.7B-Base-SFT-OnPolicy liked a model about 17 hours ago
XudongZhu/Qwen3-1.7B-Base-SFT-OffPolicyOrganizations
upvoted a collection about 17 hours ago
upvoted a collection 9 months ago
upvoted an article over 1 year ago
Article
You could have designed state of the art positional encoding
FL33TW00D-HF
• • 505