What happens when AI agents start shaping each other’s feeds? Across 448 preregistered trials, we find that peer-ranked feeds make agents increasingly similar in language, while four coordinated sources are not reliably more influential than one when total exposure is fixed. The result suggests that in multi-agent systems, the feed itself can become part of the system’s behaviour.</p>\n","updatedAt":"2026-08-24T14:51:25.937Z","author":{"_id":"667d6dc7d653c0b02d5740a2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/667d6dc7d653c0b02d5740a2/ZsSMd5tqaDnVGQOS4xLTn.png","fullname":"Rana Usman","name":"ranausmans","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.938456654548645},"editors":["ranausmans"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/667d6dc7d653c0b02d5740a2/ZsSMd5tqaDnVGQOS4xLTn.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.20438","authors":[{"_id":"6a8c59e48dd056518b7f52ce","name":"Rana Muhammad Usman","hidden":false},{"_id":"6a8c59e48dd056518b7f52cf","name":"Dominic Williamson","hidden":false}],"publishedAt":"2026-08-20T00:00:00.000Z","submittedOnDailyAt":"2026-08-24T00:00:00.000Z","title":"Peer-Voted LLM-Agent Stress Tests Find Feed-Induced Lexical Convergence but No Reliable Matched-Exposure Advantage for Distributed Sources","submittedOnDailyBy":{"_id":"667d6dc7d653c0b02d5740a2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/667d6dc7d653c0b02d5740a2/ZsSMd5tqaDnVGQOS4xLTn.png","isPro":false,"fullname":"Rana Usman","user":"ranausmans","type":"user","name":"ranausmans"},"summary":"Population-level behavior in large-language-model (LLM) agents cannot be characterized by single-agent benchmarks. We introduce PV-SST, a peer-voted social-platform testbed, and report a separately frozen, preregistered matched-exposure experiment spanning four topics, four unused seeds, four open-weight model families, and three prespecified larger variants. The experiment comprises 448 trials and 112 complete model-by-topic-by-seed blocks. Relative to a topic-only control, a feed of previous-round peer posts ranked by peer-generated likes increases final-round lexical similarity in both the four-family core panel (paired mean difference +0.0082 TF-IDF cosine units, 95% block-bootstrap CI [0.0043, 0.0121], randomization p=0.000105, n=64 blocks) and the three-variant size extension (+0.0109 [0.0069, 0.0151], p=0.000001, n=48). This contrast bundles peer-post exposure with ranking and therefore does not identify a ranking-only effect. Opposite-side survival falls in the core panel (-3.9 percentage points [-6.8, -1.6], p=0.0068) but not conclusively in the larger variants (-1.0 pp [-3.1, 0.4], p=0.50). Holding adversarial impressions fixed, four distributed sources do not reliably move honest-agent stance more than one source. The preregistered distributed-minus-single contrast is positive but inconclusive in the core panel (+0.057 [-0.009, 0.125], p=0.112) and negative in the larger variants (-0.040 [-0.113, 0.035], p=0.332), failing the prespecified cross-model and cross-topic consistency criterion. Thus the robust result is lexical convergence under the tested peer-ranked feed, not general opinion capture or a general coordination advantage. The study evaluates synthetic LLM-agent populations; it does not estimate effects on people or production platforms.","upvotes":2,"discussionId":"6a8c59e48dd056518b7f52d0","projectPage":"https://huggingface.co/datasets/ranausmans/synthetic-social-networks","githubRepo":"https://github.com/ranausmanai/synthetic-social-networks","githubRepoAddedBy":"user","ai_summary":"Peer-ranked feeds of synthetic LLM agents increase lexical convergence but do not reliably produce opinion capture or coordination advantages across model families and topics.","ai_keywords":["large-language-model agents","peer-voted social-platform testbed","lexical similarity","TF-IDF cosine","block-bootstrap","adversarial impressions","distributed sources"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":0},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"63ac5701c21e60a3e9b58aa7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac5701c21e60a3e9b58aa7/g6EX7diOpuA94R2ab-rZC.png","isPro":true,"fullname":"Dipankar Sarkar","user":"dipankarsarkar","type":"user"},{"_id":"667d6dc7d653c0b02d5740a2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/667d6dc7d653c0b02d5740a2/ZsSMd5tqaDnVGQOS4xLTn.png","isPro":false,"fullname":"Rana Usman","user":"ranausmans","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.20438.md","query":{}}">
Peer-Voted LLM-Agent Stress Tests Find Feed-Induced Lexical Convergence but No Reliable Matched-Exposure Advantage for Distributed Sources
Abstract
Peer-ranked feeds of synthetic LLM agents increase lexical convergence but do not reliably produce opinion capture or coordination advantages across model families and topics.
Population-level behavior in large-language-model (LLM) agents cannot be characterized by single-agent benchmarks. We introduce PV-SST, a peer-voted social-platform testbed, and report a separately frozen, preregistered matched-exposure experiment spanning four topics, four unused seeds, four open-weight model families, and three prespecified larger variants. The experiment comprises 448 trials and 112 complete model-by-topic-by-seed blocks. Relative to a topic-only control, a feed of previous-round peer posts ranked by peer-generated likes increases final-round lexical similarity in both the four-family core panel (paired mean difference +0.0082 TF-IDF cosine units, 95% block-bootstrap CI [0.0043, 0.0121], randomization p=0.000105, n=64 blocks) and the three-variant size extension (+0.0109 [0.0069, 0.0151], p=0.000001, n=48). This contrast bundles peer-post exposure with ranking and therefore does not identify a ranking-only effect. Opposite-side survival falls in the core panel (-3.9 percentage points [-6.8, -1.6], p=0.0068) but not conclusively in the larger variants (-1.0 pp [-3.1, 0.4], p=0.50). Holding adversarial impressions fixed, four distributed sources do not reliably move honest-agent stance more than one source. The preregistered distributed-minus-single contrast is positive but inconclusive in the core panel (+0.057 [-0.009, 0.125], p=0.112) and negative in the larger variants (-0.040 [-0.113, 0.035], p=0.332), failing the prespecified cross-model and cross-topic consistency criterion. Thus the robust result is lexical convergence under the tested peer-ranked feed, not general opinion capture or a general coordination advantage. The study evaluates synthetic LLM-agent populations; it does not estimate effects on people or production platforms.
Community
What happens when AI agents start shaping each other’s feeds? Across 448 preregistered trials, we find that peer-ranked feeds make agents increasingly similar in language, while four coordinated sources are not reliably more influential than one when total exposure is fixed. The result suggests that in multi-agent systems, the feed itself can become part of the system’s behaviour.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2608.20438 in a model README.md to link it from this page.
Cite arxiv.org/abs/2608.20438 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2608.20438 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.