small-4-8b-student-agents-multi-turn-tool-use-tasks-transferring-stronger CLM-0640mechanismsingle paperReviewed by AI
For small (4-8B) student agents on multi-turn tool-use tasks, transferring a stronger teacher's successful trajectories as flat task-level memory barely helps and can hurt, while splitting the same trajectories into a high-level workflow insight in prose plus concrete executable subtask examples raises accuracy sharply, with the subtask examples carrying most of the gain.
Capability: Using the tools it is given
Observed on
Text-based tool-use benchmarks with Python APIs or structured function calls (AppWorld, BFCL V3, ToolSandbox); students 4B-8B; teacher GPT-5-mini; memory built offline and frozen a.
Sources
- Validated by: Accuracy on AppWorld (168 tasks), BFCL V3 multi-turn base (200), ToolSandbox base (129) against zero-shot and three memory baselines (ReasoningBank, MemP, SASM), two runs averaged; gains +27.2/+11.2/+3.4 points averaged over four students, with ablation showing subtask memory adds e.g. +25.0 points over workflow alone for Qwen3-4B on AppWorld, and student-generated memory near zero-shot.. Measured on three benchmarks, four students, one main teacher, two repetitions; [truncated]
Either way, bring evidence: a link, or what you saw.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- For a same-family pair of reasoning and instruction-tuned 4B models, learning one interpolation ratio per module (attention block, FFN block, embedding, final norm) with a reward that scores correctness and ranks correct responses by length cuts average generated tokens by about 27% relative to the reasoning model while holding average accuracy, and beats fixed global-ratio merging and CMA-ES search over the same ratios.Cutting to the chase · reviewed by AI
- For LLM agents that must compose and then execute domain tool workflows, training separate adapters for the two roles splits the gains cleanly: supervised fine-tuning of the planner on reference tool sequences mainly improves workflow-order fidelity, while node-level RL on argument-match rewards mainly improves tool-argument grounding and final-answer quality, and combining both beats either on a 9B backbone.Planning a sequence of actions before acting · reviewed by AI
- When adapting a pretrained VLA manipulation policy with only a handful of target demonstrations, retrieving prior demonstrations at the level of LLM-decomposed subtasks (then reranking by behavior-feature similarity) raises success more than retrieval by visual, optical-flow, or state-action similarity, and the margin is concentrated on long-horizon composite tasks rather than single-skill pick-and-place.Turning plans into physical action · reviewed by AI
- For a domain-specific language essentially absent from pretraining corpora (FORM symbolic algebra), QLoRA fine-tuning an 8B model on a few thousand examples validated by executing them through the language's own interpreter beats frontier open-weight models of up to 756B parameters given syntax documentation in context, on execution rate and strict output match for the larger benchmarks, while general reasoning scores drop under 3 points.Generating and editing working code · reviewed by AI
- In multi-agent LLM pipelines over reasoning and code tasks with a pool of 3B-32B open-weight backbones, conditioning each per-step routing decision on a learned filtered memory rather than the full accumulated trajectory leaves accuracy about the same while cutting inference cost by roughly 40-50%, so most of the accumulated history is cost without accuracy value.Getting more from several models than from one · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: A setup where flat task-level teacher memory gives small students gains comparable to the hierarchical version, or where removing the code-bearing subtask tier leaves accuracy unchanged.