Retaining tool-use trajectories during sequential fine-tuning on API domains improves next-call prediction accuracy by 17.7 points over stripped-history training.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Trajectory Supervision for Continual Tool-Use Learning in LLMs
Retaining tool-use trajectories during sequential fine-tuning on API domains improves next-call prediction accuracy by 17.7 points over stripped-history training.