Search papers, labs, and topics across Lattice.
This paper introduces Evolution Fine-Tuning (EFT), a novel mid-training paradigm that enables large language models (LLMs) to leverage evolutionary search trajectories across multiple optimization tasks, rather than treating each task in isolation. By constructing the Finch Collection, a dataset of 156K trajectories across 371 tasks, the authors fine-tune LLMs ranging from 2B to 9B parameters, resulting in an average performance improvement of 10.22% on 22 held-out tasks compared to their base models. The findings suggest that EFT enhances cross-task generalization and can achieve state-of-the-art results when combined with test-time reinforcement learning, positioning LLMs as more effective general-purpose discovery agents.
EFT enables LLMs to evolve solutions across diverse optimization tasks, achieving over 10% performance gains and state-of-the-art results in challenging mathematical problems.
Would experience designing faster GPU kernels also help close in on a long-standing open mathematical conjecture? Large Language Models (LLMs) integrated into evolutionary search have recently produced state-of-the-art solutions on optimization tasks, including open mathematical conjectures, GPU kernel design, scientific law discovery, and combinatorial puzzles. To achieve this, prior work applied search scaffolds to one target task at a time, so every new problem is approached from scratch and the experience accumulated during search is discarded once the model finishes its attempt. This leaves the capability of iteratively evolving a solution (e.g., knowing which part to mutate and how, deciding when to backtrack) entirely in the scaffold rather than in the model itself. Whether the model itself could acquire this capability and reuse it across different tasks has been largely unexamined. To address this, we introduce Evolution Fine-Tuning (EFT), a mid-training paradigm that teaches LLMs to evolve solutions across tasks by converting evolutionary search trajectories into supervision. We construct Finch Collection, a 156K-trajectory dataset spanning 10 domains and 371 optimization tasks, and fine-tune open-source LLMs from 2B to 9B parameters. Empirically, EFT confers cross-task generalization: across 22 held-out tasks, our models surpass their base counterparts by 10.22% on average. Furthermore, when paired with test-time RL, our model matches state-of-the-art performance on two circle-packing tasks and outperforms its base-model counterpart on the Erd\H{o}s minimum-overlap problem. EFT thus serves as a"practice phase"for general-purpose discovery agents that do not solve new problems from scratch.
CMU ML