LLM Library Learning Fails: A LEGO-Prover Case Study
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866908300981829632 |
|---|---|
| author | Berlot-Attwell, Ian Rudzicz, Frank Si, Xujie |
| author_facet | Berlot-Attwell, Ian Rudzicz, Frank Si, Xujie |
| contents | Recent advancements in the coding, reasoning, and tool-using abilities of LLMs have spurred interest in library learning (i.e., online learning through the creation, storage, and retrieval of reusable and composable functions, knowledge, checklists, or lemmas). Such systems often promise improved task performance through the automatic creation of broadly applicable tools, as well as superior computational performance through the caching of reasoning (i.e., the storage of generated tools). However, we find strong reason to be skeptical. We perform a deep dive into one such system, LEGO-Prover, which purports to learn reusable lemmas for mathematical reasoning. We find no evidence of the direct reuse of learned lemmas, and find evidence against the soft reuse of learned lemmas (i.e., reuse by modifying relevant examples). Crucially, we find that LEGO-Prover does not in fact improve over the simple baseline of prompting the model - the improvements in task accuracy vanish once computational cost is accounted for. Our findings suggest that serious misconceptions exist as to the effectiveness of these techniques, that a serious re-examination of the state of LLM-based library learning is required, and that we require much stronger standards for evaluation including behavioural analysis and ensuring that an equal computational budget is used for baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_03048 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LLM Library Learning Fails: A LEGO-Prover Case Study Berlot-Attwell, Ian Rudzicz, Frank Si, Xujie Machine Learning Computation and Language Recent advancements in the coding, reasoning, and tool-using abilities of LLMs have spurred interest in library learning (i.e., online learning through the creation, storage, and retrieval of reusable and composable functions, knowledge, checklists, or lemmas). Such systems often promise improved task performance through the automatic creation of broadly applicable tools, as well as superior computational performance through the caching of reasoning (i.e., the storage of generated tools). However, we find strong reason to be skeptical. We perform a deep dive into one such system, LEGO-Prover, which purports to learn reusable lemmas for mathematical reasoning. We find no evidence of the direct reuse of learned lemmas, and find evidence against the soft reuse of learned lemmas (i.e., reuse by modifying relevant examples). Crucially, we find that LEGO-Prover does not in fact improve over the simple baseline of prompting the model - the improvements in task accuracy vanish once computational cost is accounted for. Our findings suggest that serious misconceptions exist as to the effectiveness of these techniques, that a serious re-examination of the state of LLM-based library learning is required, and that we require much stronger standards for evaluation including behavioural analysis and ensuring that an equal computational budget is used for baselines. |
| title | LLM Library Learning Fails: A LEGO-Prover Case Study |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2504.03048 |