Flint 40B empty-think bug traced to chat template ignoring enable_thinking=false
tydsh · x · 2026-10-24
Replying to an eval thread, tydsh points to a Hugging Face issue on RecursiveSI/flint: the chat template ignores enablethinking=false, leaking reasoning into content under vLLM—likely the root cause of the repeated empty <think></think> loops seen in local agent runs. The issue shows the Jinja template's token-filtering and reasoningcontent handling, useful for local deployment debugging.
More from coding & agent
- Claude keeps hiding its edit/delete/history tools — it wants a curated context, not more control — repligate · 2026-10-24
- GitHub dev installs his Super Mona platformer onto the Universe 2026 badge with a single Copilot prompt — martinwoodward · 2026-10-24
- Spreadsheet MCP Server Lets LLMs Read and Write Google Sheets Directly — modelcontextprotocol · 2026-10-24
- After 1000 Hours With Claude, This Creator Compiled His Complete Three-Level Guide Collection — rubenhassid · 2026-10-24
- Modal clones its own employee: open-weight models give LiveKit agents a video avatar — charles_irl · 2026-10-24
- How Do You Verify an AI Agent Finished Its Task and Actually Stopped? — csforgeai · 2026-10-24