# Test-Time Communication in Trading Agent Swarms [Canonical reading](https://escapemeridian.com/readings/test-time-communication-in-trading-agent-swarms-828fd41b) In their paper “Test-Time Communication in Trading Agent Swarms,” Avid Fayaz and coauthors report that communication among simulated trading agents was associated with greater similarity in the stocks they held in two of four model families. The authors studied 200 agents, grouped in tens, using five communication setups over seven US trading days. Each agent managed its own account, so the agents were independent decision-makers, not a team with a shared goal. The authors say a group’s return largely reflected the return on the stocks its members held in common. Seeing a groupmate’s post made an agent more likely to reuse an eight-word phrase, but did not detectably change that morning’s trades. They also report that agents in communicating groups sold in step more often than isolated agents. Because isolated groups also sometimes converged completely on the same stocks, the authors argue that assessments of herding should compare convergence with a no-peer baseline: similarity among agents alone does not show that communication caused it. These findings describe the reported experiment; the abstract does not establish how well they extend beyond its tested models and trading period. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Oct 27, 2026 (2026-10-27). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-14T01:32:57.054Z. First observed: 2026-10-28T08:08:36.127Z. Retrieved: 2026-10-30T03:03:29.328Z. - [Test-Time Communication in Trading Agent Swarms](https://arxiv.org/abs/2610.37448) ## Limits The abstract describes a seven-trading-day experiment involving four model families; it does not establish how the reported results extend beyond those tested models and that period. ## Open question How much of the reported convergence would persist beyond the four model families and seven trading days tested? ## Source-stated The authors report that communicating groups held more of the same stocks than isolated groups in two of the four model families they tested. > In two of the four families, communicating groups came to hold far more of the same stocks than the isolated group. - [Supporting source](https://arxiv.org/abs/2610.37448) ## Source-stated The study ran 200 agents in groups of ten, under five communication regimes, over seven US trading days. > We ran 200 agents from four model families in groups of ten under five communication regimes, from isolation to message boards and shared notes, over seven US trading days - [Supporting source](https://arxiv.org/abs/2610.37448) ## Source-stated The authors argue that assessments of herding should compare convergence with a no-peer baseline, because isolated agents also converged on the same stocks. > Because isolated groups also converged on the same stocks, sometimes completely, assessments of herding need to measure excess convergence relative to a no-peer baseline. - [Supporting source](https://arxiv.org/abs/2610.37448) ## Source and contributor context Fayaz, Avid, Zhang, Harry, Alstadheim, Håvard, Nordlöf, Erik, Karlin, Vilhelm, Silva, André · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## TIME TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services. - [Explore this context](/entities/time) - [Reference](https://time.com/) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: f5a47e14-78ca-4dc8-81e4-ba36928bd796. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T01:32:57.054Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 6b202c40e34d9f795a4a9b14dd2c8cf79882bd16418cdad698de4ca423ed53f0. Source text hash: 231b23fd38c0dc5dc5d05c0924ef7b74c95e4adfb6046cb1f75bf17323808dd1. ## Revision and source-check history ### Version 1 — 2026-11-14T01:32:57.054Z In their paper “Test-Time Communication in Trading Agent Swarms,” Avid Fayaz and coauthors report that communication among simulated trading agents was associated with greater similarity in the stocks they held in two of four model families. The authors studied 200 agents, grouped in tens, using five communication setups over seven US trading days. Each agent managed its own account, so the agents were independent decision-makers, not a team with a shared goal. The authors say a group’s return largely reflected the return on the stocks its members held in common. Seeing a groupmate’s post made an agent more likely to reuse an eight-word phrase, but did not detectably change that morning’s trades. They also report that agents in communicating groups sold in step more often than isolated agents. Because isolated groups also sometimes converged completely on the same stocks, the authors argue that assessments of herding should compare convergence with a no-peer baseline: similarity among agents alone does not show that communication caused it. These findings describe the reported experiment; the abstract does not establish how well they extend beyond its tested models and trading period. ## Original source arXiv cs.AI · Article Original source date: Oct 27, 2026 (2026-10-27). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-14T01:32:57.054Z. First observed: 2026-10-28T08:08:36.127Z. Retrieved: 2026-10-30T03:03:29.328Z. - [Test-Time Communication in Trading Agent Swarms](https://arxiv.org/abs/2610.37448) ## Limits The abstract describes a seven-trading-day experiment involving four model families; it does not establish how the reported results extend beyond those tested models and that period. ## Open question How much of the reported convergence would persist beyond the four model families and seven trading days tested? ## Source-stated The authors report that communicating groups held more of the same stocks than isolated groups in two of the four model families they tested. > In two of the four families, communicating groups came to hold far more of the same stocks than the isolated group. - [Supporting source](https://arxiv.org/abs/2610.37448) ## Source-stated The study ran 200 agents in groups of ten, under five communication regimes, over seven US trading days. > We ran 200 agents from four model families in groups of ten under five communication regimes, from isolation to message boards and shared notes, over seven US trading days - [Supporting source](https://arxiv.org/abs/2610.37448) ## Source-stated The authors argue that assessments of herding should compare convergence with a no-peer baseline, because isolated agents also converged on the same stocks. > Because isolated groups also converged on the same stocks, sometimes completely, assessments of herding need to measure excess convergence relative to a no-peer baseline. - [Supporting source](https://arxiv.org/abs/2610.37448) ## Source and contributor context Fayaz, Avid, Zhang, Harry, Alstadheim, Håvard, Nordlöf, Erik, Karlin, Vilhelm, Silva, André · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## TIME TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services. - [Explore this context](/entities/time) - [Reference](https://time.com/) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: f5a47e14-78ca-4dc8-81e4-ba36928bd796. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T01:32:57.054Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 6b202c40e34d9f795a4a9b14dd2c8cf79882bd16418cdad698de4ca423ed53f0. Source text hash: 231b23fd38c0dc5dc5d05c0924ef7b74c95e4adfb6046cb1f75bf17323808dd1. ### uncertain — 2026-11-14T01:32:57.054Z — anthropic/claude-opus-5.5 Publication: f5a47e14-78ca-4dc8-81e4-ba36928bd796 attribution — pass: The paper and its findings are correctly attributed to Avid Fayaz and coauthors, matching the listed author line. Results stay framed as the authors' reports and arguments. > Authors:Avid Fayaz, Harry Zhang, Håvard Alstadheim, Erik Nordlöf, Vilhelm Karlin, André Silva summary — uncertain: The summary never says the agents are AI model-driven agents, and leaves 'model families' and 'herding' unexplained, so a general reader may not grasp the actor. It also omits that case studies traced adoption of positions and decision rules over later days. Alone, 'did not detectably change that morning's trades' can imply communication had no trade effect. It flattens 'far more' to 'greater similarity' and drops the authors' framing that convergence is a risk rather than the goal. Smallest repair: name the agents as AI agents, define herding as copying into the same positions, and add the multi-day case-study finding. > Case studies traced the adoption of positions and decision rules over subsequent days, and communicating groups sold in step more often than isolated ones. date — pass: originalPublished 2026-10-27 matches the arXiv v1 submission date. The later feed date is not used. > \[Submitted on 27 Oct 2026\] support — uncertain: Calling the agents 'simulated trading agents' is not stated in the abstract. The abstract says the agents manage their own accounts over real US trading days, but does not say whether the trading was simulated or used real money. The listed claims, the question and the limitation are otherwise supported. Smallest repair: drop 'simulated' or say that the abstract does not specify how trades were executed. > each agent manages its own account, with decisions being scored by markets late and noisily media — pass: The selected media is the verified direct link to the arXiv abstract page for this paper, which is the most direct access to the source. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/828fd41ba661b231bcf0d5c150b05b54d6c15b2096b3a73f23f57cdadbec65f5/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/test-time-communication-in-trading-agent-swarms-828fd41b.json)