A pipeline that synthesizes verifiable tasks from tool trajectories to train and evaluate LLM agents on ambiguous, changing, and infeasible user intents.
State in `task_instructions` that the user wants to know or act based on that amount
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Trajectory2Task: Training Robust Tool-Calling Agents with Synthesized Yet Verifiable Data for Complex User Intents
A pipeline that synthesizes verifiable tasks from tool trajectories to train and evaluate LLM agents on ambiguous, changing, and infeasible user intents.