Chinese contentSchemaShield-Lite：无需重训即可修复 LLM 智能体的接口漂移问题
做智能体工具调用的朋友看这篇，工具接口一改模型就崩，SchemaShield-Lite 不用重训就能在 BFCL-200 上把准确率从 0.58 拉到 0.805。
#BFCL
5 stories · 1 in 7 days · 3 in 30 days
Stories, products and papers tagged "BFCL", newest first.
10月19日
10月15日
Chinese contentSeTPO 训练方法让小语言模型更擅长调用外部工具
小模型学用工具的老难题有了新解法：SeTPO 用双向 KL 同时管住探索和模仿，BFCL v4 上跑赢了蒸馏和纯 RL，做端侧模型的可以看看。
9月28日
Critical-State RL Improves Multi-Turn Tool Use in Salesforce Paper
The insight is that most reward variation in multi-turn RL comes from downstream noise, so credit assignment should target the pivotal call rather than the full trajectory. Researchers and engineers building tool-using agents can apply nested sampling and contextual-bandit updates to concentrate training on decision-critical states. The BFCL v4 result offers a concrete benchmark for comparing against trajectory-level reward approaches.
9月22日
Chinese contentCritical-State RL：定位多轮工具调用中值得训练的状态
多轮工具调用失败，不知道该练哪一步？这篇论文给出诊断方法，BFCL v4 缺失函数任务提升 14 个百分点。
8月31日