Talentrl improves task success by guiding skill choice with strategy
TalentRL: Strategy-Guided Reinforcement Learning for Hierarchical Skill Routing
Machine Learning
Summary
Many AI systems that answer questions use a big list of skills but don’t decide which skills to use based on a bigger strategy. The authors created TalentRL, which first picks a general strategy, then finds the best skills for that strategy to solve a task. They trained this method using a clever way to figure out how much each strategy choice helped results. TalentRL performed better than other methods on multiple question-answering tasks and AI challenges involving interaction.
What this means in practice
- •For ai application developers: Improve multi-step or complex AI question-answering systems by guiding skill selection with strategies to boost accuracy and success rates.
- •For interactive agent builders: Enhance agents that perform tasks in simulated or web environments by better routing between skills according to high-level strategies.
Authors
Huancheng Chen, Xiaodi Sun, Zhaoqiong Huang, Michelle Chang, Jingwen Lu
Abstract
Natural-language skills enable large language model (LLM) agents to reuse procedures across tasks, yet most systems retrieve individual skills from a flat SkillBank without explicitly modeling the high-level strategy that governs their selection and use. We introduce TalentRL, a strategy-guided reinforcement learning framework for hierarchical routing over procedural memory. Coarse-grained talents represent recurring strategies for skill use, while fine-grained skills provide executable procedures. For each task, the policy model selects a talent that guides execution and softly prioritizes its associated skills during similarity-based retrieval. However, since task outcomes depend jointly on talent selection, retrieved skill context, and downstream execution, terminal rewards alone cannot isolate the contribution of the routing decision. TalentRL therefore trains talent routing with reward contrasts between rollouts of the same task under matched interventions on the talent-conditioned context, and optimizes downstream execution with group-relative advantages. Across seven open-domain and multi-hop question-answering benchmarks, talent-aware supervised fine-tuning improves average success rate by 7.53%, 6.71%, and 9.21% over the corresponding zero-shot baselines for Qwen2.5-7B, Qwen3-8B, and Llama-3.1-8B, respectively. Reinforcement learning further improves these SFT checkpoints by 7.58% to 9.32%, and TalentRL achieves the highest average success rate among all compared methods on all three backbones. On the agentic benchmarks ALFWorld and WebShop, TalentRL achieves the highest ALFWorld success rate (99.3%) and the highest WebShop score (90.7) among all compared methods.