Paper Detail

ProgAgent: Learning Tool Orchestration through Programmatic Training

Hao Li, Keqing He, Longtao Zheng, Junlong Li, Weihao Zeng, Qian Liu, Yuzhen Huang

huggingface Score 14.0

Published 2026-10-10 · First seen 2026-10-15

General AI

Abstract

Large language models (LLMs) are increasingly expected to orchestrate diverse tools over long-horizon tasks. Existing agents typically use Direct Tool Calling (DTC), where the model invokes tools sequentially, and tool responses accumulate in the model context, which can lead to context rot and orchestration breakdown. In contrast, Programmatic Tool Calling (PTC) invokes tools through executable code and stores intermediate results in runtime variables rather than appending them to the context. Prior work has explored PTC mainly as an inference-time optimization, but its benefits for agentic training remain unclear, and suitable verifiable training tasks are scarce. In this work, we introduce ProgAgent, a full PTC training stack, comprising a pipeline for synthesizing verifiable long-horizon tasks, a lightweight PTC runtime, and a two-stage training recipe. Using only 215 tasks instantiated at four environment scales, ProgAgent yields higher pass rates than the evaluated baselines of comparable size across four benchmarks. On LOCA-bench, ProgAgent-14B achieves a 30.7% pass rate at a 64k-token environment state, where these baselines score zero. Cross-harness experiments show that PTC-trained models outperform DTC-trained models under both evaluation interfaces, suggesting that the benefits of PTC training extend beyond the interface.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
now
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@misc{li2026progagent,
  title = {ProgAgent: Learning Tool Orchestration through Programmatic Training},
  author = {Hao Li and Keqing He and Longtao Zheng and Junlong Li and Weihao Zeng and Qian Liu and Yuzhen Huang},
  year = {2026},
  abstract = {Large language models (LLMs) are increasingly expected to orchestrate diverse tools over long-horizon tasks. Existing agents typically use Direct Tool Calling (DTC), where the model invokes tools sequentially, and tool responses accumulate in the model context, which can lead to context rot and orchestration breakdown. In contrast, Programmatic Tool Calling (PTC) invokes tools through executable code and stores intermediate results in runtime variables rather than appending them to the context. },
  url = {https://huggingface.co/papers/2610.14774},
  keywords = {Direct Tool Calling, Programmatic Tool Calling, ProgAgent, long-horizon tasks, verifiable training, two-stage training, runtime variables, context rot, code available, huggingface daily},
  eprint = {2610.14774},
  archiveprefix = {arXiv},
}

Metadata

{}