ProtAgent：8B 小模型通过交互级信用分配后训练超越 DeepSeek-V3.2-Thinking
ProtAgent: Post-Training Compact Protein Tool Agents with Interaction-Level Credit Assignment
8B 模型后训练后在蛋白质注释任务上干翻了 671B 的 DeepSeek-V3.2-Thinking，论文里逐步级奖励分配的对比实验讲得很细，做智能体训练的可以看看。
ProtAgent 是一个两阶段后训练流程，先用 SFT 让模型学会工具接口，再用 on-policy REINFORCE 优化工具交互策略。其核心是步骤级回报机制：中间工具调用获得动作奖励，最终答案获得终端奖励。在固定闭环协议下，ProtAgent-8B 的标注覆盖率成功率达 61.00%，超过 DeepSeek-V3.2-Thinking 671B 的 41.60%，与最强前沿系统 60.80% 基本持平。三种子对比显示，步骤级回报下平均成功率 61.40%，而片段级回报只有 42.70%，说明信用分配粒度是关键变量。
ProtAgent: Post-Training Compact Protein Tool Agents with Interaction-Level Credit Assignment
Protein annotation agents must choose suitable bioinformatics tools, interpret their outputs, and use the resulting evidence in the final answer. Most existing systems rely on frontier-scale models. We ask whether a compact model can instead learn this behavior through post-training with real tool interaction. We present ProtAgent, a two-stage pipeline that uses supervised fine-tuning (SFT) to learn the tool interface and standard on-policy REINFORCE to improve the interaction policy. Intermediate tool calls receive action rewards, and the final response receives a terminal answer reward. Step-level returns assign these rewards to the responses that produced them and to preceding actions. The episode-level baseline instead assigns one episode return to every action. On a fixed closed-loop protocol with real bioinformatics tools, ProtAgent-8B achieves 61.00% success under the annotation-coverage metric, compared with 41.60% for DeepSeek-V3.2-Thinking 671B and 60.80% for the strongest evaluated frontier system. In the three-seed comparison, mean success is 42.70% with episode-level returns and 61.40% with step-level returns under identity return normalization; the corresponding group-relative results are 36.50% and 58.20%. The metric measures recovery of reference annotations under this protocol; it does not by itself establish biological correctness or complete evidence grounding.
