Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
EvoIn decouples agent decision-making from persistent scaffolding by evolving procedural heuristics in a temporary harness, using them to generate high-performing execution traces, and rewriting those traces to express the harness logic as internal model reasoning. Fine-tuning models on these rewritten traces improves agent pass rates by 10.9 points in-domain and 9.2 points out-of-domain while successfully transferring across model families. This approach enables models to natively internalize meta-deliberation strategies—such as inspecting document lengths before choosing between reading and searching—without requiring complex orchestration logic at inference time.
External agent scaffolding can be discarded after training by distilling evolved decision-making procedures directly into a model's native reasoning traces.
Recent work has explored improving agents by jointly evolving their harnesses and models, but often takes a ''potpourri'' approach that bundles together new tools, new decision-making procedures, and model adaptation to the evolved harness under a single notion of agent improvement. In this paper, we instead investigate how agents can improve their decision-making procedures. In particular, we propose EvoIn, an agent fine-tuning framework that bridges evolution and internalization. EvoIn first analyzes agent execution traces to evolve and validate new decision-making procedures by temporarily instantiating them in the harness. The validated procedures guide the agent to generate improved reasoning traces. These traces are then rewritten into self-contained reasoning traces, removing explicit references to harness instructions while expressing the induced decision logic as the model's own reasoning. Finally, EvoIn fine-tunes the model on the rewritten traces, internalizing these procedures so that the improved decision-making persists without the evolved harness at inference time. We evaluate EvoIn on diverse benchmarks and find that it consistently enables agents to learn stronger decision-making procedures, raising the pass rate by 10.9 points in-domain and by 9.2 points out-of-domain. Results further show that the internalized decision procedures generalize to unseen tasks. Case studies show that agents can learn to decide how to solve a task before solving it, for example by checking a document's length to choose between reading it in full and searching it. EvoIn is also broadly applicable, showing consistent improvements on another model family.