Qwen3-4B checkpoints whose chat template always preserves <think> reasoning across multi-turn history, with no nonthinking mode.
Ethan Ewer
eewer
AI & ML interests
None yet
Recent Activity
updated a dataset 29 days ago
eewer/terminal-midtraining-trace-scores published a dataset 29 days ago
eewer/terminal-midtraining-trace-scores updated a model about 2 months ago
eewer/nemotron3-nano-v0-11-gbs12-lr1e5-base-iter15320Organizations
None yet