@maximelabonne: New training technique to reduce doom loops! We applied it to LFM2.5-2.6B (SFT checkpoint) Qwen3.5-4B. By reducing doom…
Summary
Liquid AI releases Antidoom, an open-source method to reduce doom loops in reasoning models, applied to LFM2.5-2.6B and Qwen3.5-4B, significantly lowering doom-loop rates and improving eval scores.
View Cached Full Text
Cached at: 07/07/26, 02:19 PM
New training technique to reduce doom loops!
We applied it to LFM2.5-2.6B (SFT checkpoint) Qwen3.5-4B.
By reducing doom loops, it also improves downstream evals.
We open-source the training code and training dataset on @huggingface
Liquid AI (@liquidai): Today we release Antidoom, an open-source method that removes a common failure mode in reasoning models: the doom loop.
Doom-loop rates before and after, with eval scores up across the board:
> Early LFM2.5-2.6B checkpoint: 10.2% → 1.4%
> Qwen3.5-4B: 22.9% → 1% (greedy
Similar Articles
@LiorOnAI: An open-source fix for one of the most common reasoning model failure modes. One of the biggest AI trends this year isn…
Liquid AI releases Antidoom, an open-source method that fine-tunes reasoning models to break repetitive token loops (doom loops), reducing failure rates from ~23% to 1% on Qwen3.5-4B without retraining or RL.
Reducing Doom Loops with Final Token Preference Optimization
Liquid AI introduces Antidoom, a method using Final Token Preference Optimization to reduce repetitive doom loops in small reasoning models during inference, cutting loop rates from 10.2% to 1.4% while improving evaluation scores.
@helloiamleonie: Working with the @liquidai team on these engineering blogs is just so much fun! Here's what we've been working on: Reas…
The post explains doom loops in reasoning models where the model repeats tokens like 'Wait' until the context fills up, and introduces FTPO (Final Token Preference Optimization) as a training-time fix. The associated Antidoom tool reduces doom loop rates significantly (e.g., from 22.9% to 1% on Qwen3.5-4B).
@Raman_bansal_: If you’ve trained an LLM, you’ve may have seen doom loop, in which a LLM endlessly repeats the same token or sentence, …
A Substack article explains the 'doom loop' problem in LLMs where models repeat tokens endlessly, and introduces Final Token Preference Optimization (FTPO) from Liquid AI as a method to detect and fix such loops during fine-tuning.
I created an LLM post-training method called RPS. Preliminary results show that it improved Qwen3-8b's program synthesis reliability. [R]
RPS is a two-stage LLM post-training method inspired by neuroscience, combining curriculum learning with learning rate decay. Preliminary results show improved program synthesis reliability on Qwen3-8b compared to equal learning rate training.