4b-8b-open-base-models-mid-trained-tool-use-grounding-oriented-supervision CLM-0529mechanismsingle paperReviewed by AI
For 4B-8B open base models mid-trained for tool use, grounding-oriented supervision (documentation/code turned into QA) and execution-oriented supervision (executable trajectories from real APIs and MCP skills) help different things, and only the combination improves everything: trajectories alone give the bigger function-calling gain while grounding alone transfers better to interactive and unseen-tool benchmarks, and each branch alone scores below no mid-training on MCP-Universe.
Capability: Using the tools it is given
Observed on
Qwen3-4B-Base with a fixed downstream SFT recipe on a 100K tool-use subset; evaluated on BFCLv3, tau2-Bench and MCP-Universe; 20.3B-token mid-training budget..
Sources
- Validated by: Additive ablation at 4B with the SFT recipe held fixed: no mid-training 39.73% BFCLv3 overall, 8.54% tau2 Pass@1, 13.20 MCP score; raw processed sources only 42.30/7.30/12.20; +native trajectories only 47.59/4.23/6.80; +context-grounded only 44.66/8.99/8.46; full mixture 50.25/12.23/18.66. Matched-budget Dolmino-20BT baseline reached 43.10 BFCLv3 but 5.41 MCP score.. Measured in a single additive ablation at one model scale (4B) with one fixed SFT recipe; [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
- When fine-tuning a 4B code model, training on auxiliary targets that name each program region's correctness obligations (derived from aligning an AST implementation graph with Lean proof-flow traces) raises solved counts over code-only SFT, but the auxiliary-target stage alone shifts the model off the Python-only output distribution and must be followed by a code-only consolidation stage to avoid losing broad function-level ability.Generating and editing working code · 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
- 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 7-8B open-weight instruct models are fully fine-tuned on a single domain dataset, confining gradient updates to FFN neurons that activate stably across the training set and adding an MSE loss that pulls activations back toward pre-computed activation anchors retains far more general-benchmark ability (MMLU, BBH, TyDiQA) than full-parameter fine-tuning, by about 9.9 points on average, while also raising in-domain scores.Retaining skills and safety after fine-tuning · reviewed by AI
- For small open-weight backbones (Qwen3-4B) driving long-horizon remote-sensing tool chains, inserting a step-level verifier that checks each intermediate tool output against schema, sub-goal and domain constraints and triggers replanning is the single largest architectural gain, while hierarchical decomposition without domain-aligned tool grouping adds little and can lower final accuracy relative to a flat agent with the same verifier.Completing tasks that take many steps or many hours · 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: An ablation at matched token budget where one branch alone matches or beats the combined mixture across function-calling, interactive and unseen-tool benchmarks, or where neither branch shows the reported asymmetry.