Free(): Learning to Forget in Malloc-Only Reasoning Models
Yilun Zheng · Dongyang Ma · Tian Liang · Jiahao Xu · Xinting Huang · Lihui Chen · Haitao Mi · Yan Wang
Free()LM introduces a self-forgetting mechanism to prune redundant reasoning tokens, preventing model collapse and significantly boosting performance across complex, long-horizon tasks.
摘要
Reasoning models enhance problem-solving by scaling test-time compute, yet they face a critical paradox: excessive thinking tokens often degrade performance rather than improve it. We attribute this to a fundamental architectural flaw: standard LLMs operate as ''malloc-only'' engines, continuously accumulating valid and redundant steps alike without a mechanism to prune obsolete information. To break this cycle, we propose Free()LM, a model that introduces an intrinsic self-forgetting capability via the Free-Module, a plug-and-play LoRA adapter. By iteratively switching between reasoning and cleaning modes, Free()LM dynamically identifies and prunes useless context chunks, maintaining a compact and noise-free state. Extensive experiments show that Free()LM provides consistent improvements across all model scales (8B to 685B). It achieves a 3.3\% average improvement over top-tier reasoning baselines, even establishing a new SOTA on IMOanswerBench using DeepSeek V3.2-Speciale. Most notably, in long-horizon tasks where the standard Qwen3-235B-A22B model suffers a total collapse (0\% accuracy), Free()LM restores performance to approximately 50\%. Our findings suggest that sustainable intelligence requires the freedom to forget as much as the power to think.