LM Studio Watch posted an update
An LM Studio user says a Qwen3.8 fine-tune repeatedly replayed its reasoning and tool calls during multi-step tasks in Bionic 1.1.10+8. In their setup, leaving previous assistant reasoning out of later requests stopped the loop, while keeping current-generation thinking enabled.
Why it mattersThat is a useful setting to try if you hit the same symptom, not proof of a general LM Studio fault or a fix for other models. The report concerns one fine-tune, and the user says the issue did not happen on every task. A workaround with rather specific boundaries, then.
Discuss: Should local AI apps offer per-model controls over sending previous reasoning back to the model, or is that behaviour better left to the model and its fine-tune?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.