Federation Is Nearly Free, Reasoning Is Not: Tradeoffs for AI Co-Scientists in Protein Characterization Workflows

Summary

A controlled ablation on a production scientific-agent platform varying federation topology, harness type (classic RL versus LLM-driven), language model, and prompt expertise on a verifiable protein-characterization task, stratified by protein novelty. Model choice dominates prediction quality (Opus ~92-94% versus o4-mini ~40-50%) while federation topology imposes a negligible penalty, and a PPO policy reaches 88% at zero token cost with fastest latency and perfect consistency but yields no reasoning trace. Two things make it relevant: it is a rival finding that topology is second-order to model capability, a threat the ISR paper must confront head-on, and it renders the traceability-accuracy tradeoff concrete by showing the cheapest accurate configuration is the one with no auditable justification.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-09-07 literature sweep as a medium-priority candidate for the Architecture as Epistemology (ISR) paper (angle: organization theory applied to AI agents; MAS failure and cost analyses). Verified against: Verified via arXiv API metadata record on 2026-09-07: full abstract, 4 authors, v1, published 2026-08-25, primary cs.AI (also cs.MA, q-bio.QM), 24 pages.

Critique / open questions

Queued — not yet assessed.

Key quotes

Natural language driven autonomous co-scientist workflows involve a fundamental trade-off between flexibility and reasoning at the expense of determinism, reproducibility, and observability. (abstract)

Such agents increasingly must communicate across institutional boundaries, where federation topology can shape latency and cost. (abstract)