Paper Detail

SciOrch: Learning to Orchestrate Expert LLMs for Solving Frontier Multimodal Scientific Reasoning Tasks

Jingru Guo, Xiangyuan Xue, Lian Zhang, Wanghan Xu, Siki Chen, Philip Torr, Wanli Ouyang, Lei Bai, Zhenfei Yin

huggingface Score 19.5

Published 2026-06-14 · First seen 2026-06-18

General AI

Abstract

Frontier scientific reasoning remains a major challenge for large language models (LLMs), where even the strongest commercial systems fall short of expert-level performance. A closer look at model behavior reveals substantial complementarity that single-model evaluation hides: different frontier models excel on different question types, and no single model captures the full picture. We present SciOrch, a framework that trains a lightweight 8B model to orchestrate frontier LLMs for scientific reasoning. The orchestrator decomposes each question, delegates sub-problems to selected commercial models through API calls, and synthesizes a final answer. Training such an orchestrator is fundamentally harder than conventional agentic RL: each action triggers an API call that is expensive in both dollar cost and latency, making standard online rollouts infeasible. We address this with MCTS-based approach, producing diverse orchestration trajectories, extracting per-node single-turn samples, and optimizing the orchestrator with GRPO-style training. On a 240-question test set spanning SGI-Reasoning and Scientists' First Exam, SciOrch reaches 56.66% average accuracy, outperforming the strongest single commercial model by 3.74% and the strongest multi-agent baseline by 3.33%. It also attains the best accuracy on both SGI and SFE with less than half the API cost of typical multi-agent methods.

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{guo2026sciorch,
  title = {SciOrch: Learning to Orchestrate Expert LLMs for Solving Frontier Multimodal Scientific Reasoning Tasks},
  author = {Jingru Guo and Xiangyuan Xue and Lian Zhang and Wanghan Xu and Siki Chen and Philip Torr and Wanli Ouyang and Lei Bai and Zhenfei Yin},
  year = {2026},
  abstract = {Frontier scientific reasoning remains a major challenge for large language models (LLMs), where even the strongest commercial systems fall short of expert-level performance. A closer look at model behavior reveals substantial complementarity that single-model evaluation hides: different frontier models excel on different question types, and no single model captures the full picture. We present SciOrch, a framework that trains a lightweight 8B model to orchestrate frontier LLMs for scientific rea},
  url = {https://huggingface.co/papers/2606.15872},
  keywords = {large language models, scientific reasoning, frontier models, orchestrator model, API calls, MCTS-based approach, GRPO-style training, multi-agent baseline, SGI-Reasoning, Scientists' First Exam, code available, huggingface daily},
  eprint = {2606.15872},
  archiveprefix = {arXiv},
}

Metadata

{}