We find a surprising failure mode, <strong>forecast collapse</strong>, of Time Series Foundation Models (TSFMs). Despite their strong performance across general time-series benchmarks, we observe that TSFMs can generate overly smooth or near-constant forecasts that fail to reflect the heterogeneous dynamics of the underlying multivariate series.</p>\n","updatedAt":"2026-08-17T05:39:02.155Z","author":{"_id":"63e8882e4577a86987ad91c2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/CmgkIiJ9Unu1HlPtp7TqF.png","fullname":"Shu","name":"Shuwan","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8607839941978455},"editors":["Shuwan"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/CmgkIiJ9Unu1HlPtp7TqF.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.14106","authors":[{"_id":"6a829c93b601d59c652814cd","name":"Shu Wan","hidden":false},{"_id":"6a829c93b601d59c652814ce","name":"Miles Ma","hidden":false},{"_id":"6a829c93b601d59c652814cf","name":"Hank Zhu","hidden":false},{"_id":"6a829c93b601d59c652814d0","name":"Guangqi Liu","hidden":false},{"_id":"6a829c93b601d59c652814d1","name":"Stephen Wang","hidden":false},{"_id":"6a829c93b601d59c652814d2","name":"Qingsong Wen","hidden":false},{"_id":"6a829c93b601d59c652814d3","name":"Huan Liu","hidden":false}],"publishedAt":"2026-08-14T00:00:00.000Z","submittedOnDailyAt":"2026-08-17T00:00:00.000Z","title":"Forecast Collapse in Time-Series Foundation Models","submittedOnDailyBy":{"_id":"63e8882e4577a86987ad91c2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/CmgkIiJ9Unu1HlPtp7TqF.png","isPro":false,"fullname":"Shu","user":"Shuwan","type":"user","name":"Shuwan"},"summary":"When forecasting hourly returns for 1,000 US equities, we observe an unexpected phenomenon: predictions become nearly flat and show poor stock ranking, as measured by cross-sectional correlation. We call this forecast collapse. Surprisingly, the phenomenon largely disappears when forecasting trading volume under the same setting. We investigate forecast collapse across time-series foundation models (TSFMs), twelve deep-learning forecasting models, and 97 public benchmark configurations, and find that it is closely tied to target predictability. We identify two distinct reasons behind it: low predictability limits the amplitude of calibrated point forecasts, while per-series objectives leave cross-series structure unidentified. These findings reveal a calibration-ranking tradeoff: optimizing squared error leads to flat predictions, whereas directly optimizing cross-sectional correlation improves ranking but can inflate forecast amplitude by more than an order of magnitude. To address this tradeoff, we introduce CalibRank, a simple objective that balances calibration and ranking. On Finance1K, CalibRank nearly triples cross-sectional correlation while keeping amplitude close to the target, and improves correlation on all tested models. Our results reveal a blind spot in conventional time-series evaluation: per-series metrics can hide failures in cross-series structure needed by downstream decisions.","upvotes":7,"discussionId":"6a829c94b601d59c652814d4","ai_summary":"Forecast collapse in hourly equity return prediction stems from low predictability and per-series objectives, and the proposed CalibRank objective balances calibration and ranking to restore cross-sectional structure.","ai_keywords":["forecast collapse","time-series foundation models","cross-sectional correlation","calibration-ranking tradeoff","CalibRank","per-series objectives"],"ai_summary_model":"thinkingmachines/Inkling-Small","organization":{"_id":"6a396f6821954f7afc3136b4","name":"abel-lab","fullname":"Abel","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/63e8882e4577a86987ad91c2/K7kvaiiKEV6YGyBNLSO6S.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"63e8882e4577a86987ad91c2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/CmgkIiJ9Unu1HlPtp7TqF.png","isPro":false,"fullname":"Shu","user":"Shuwan","type":"user"},{"_id":"6a3b6a44d3cae0e13ea74b55","avatarUrl":"/avatars/c2dc25e9aaa1c991c4a279adc900d5a4.svg","isPro":false,"fullname":"Rui Zhu","user":"RuiLGTM","type":"user"},{"_id":"6458967ad1d994ec81053829","avatarUrl":"/avatars/8f1a0df9f7c19b9f3cf285a49f59806a.svg","isPro":false,"fullname":"chen si","user":"exen3995","type":"user"},{"_id":"66613f260d1be84ebe820fdb","avatarUrl":"/avatars/da111c89af0774a143c6ee3b62364dae.svg","isPro":false,"fullname":"Zhu","user":"HankHHH","type":"user"},{"_id":"6a397c7249de0be3d22c7a13","avatarUrl":"/avatars/b47e2a3e6b637de99880c744e79734b0.svg","isPro":false,"fullname":"Stephen Wang","user":"cauchyabel","type":"user"},{"_id":"6a82a2ddefdaaaa68ae358ce","avatarUrl":"/avatars/33e4ef04b9b69853793c676b0c8aef76.svg","isPro":false,"fullname":"elisa qian","user":"elqian","type":"user"},{"_id":"63ac5701c21e60a3e9b58aa7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac5701c21e60a3e9b58aa7/g6EX7diOpuA94R2ab-rZC.png","isPro":true,"fullname":"Dipankar Sarkar","user":"dipankarsarkar","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"6a396f6821954f7afc3136b4","name":"abel-lab","fullname":"Abel","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/63e8882e4577a86987ad91c2/K7kvaiiKEV6YGyBNLSO6S.png"},"query":{}}">

Cite arxiv.org/abs/2608.14106 in a model README.md to link it from this page.

Cite arxiv.org/abs/2608.14106 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.