Hugging Face Daily Papers · May 22, 2026 · 3 min read

TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Like Read original ↗

<a href=\"https://github.com/HotTricker/TransitLM.git\" rel=\"nofollow\">https://github.com/HotTricker/TransitLM.git</a></p>\n","updatedAt":"2026-05-22T07:38:47.264Z","author":{"_id":"66d255e3947594430c723ff6","avatarUrl":"/avatars/c56e4792332a01bf34085a75ee64916e.svg","fullname":"xiaochonglinghu","name":"xiaochonglinghu","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":9,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5978816747665405},"editors":["xiaochonglinghu"],"editorAvatarUrls":["/avatars/c56e4792332a01bf34085a75ee64916e.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.22355","authors":[{"_id":"6a0fc10ea53a61ce2e422c9a","name":"Hanyu Guo","hidden":false},{"_id":"6a0fc10ea53a61ce2e422c9b","name":"Jiedong Yang","hidden":false},{"_id":"6a0fc10ea53a61ce2e422c9c","name":"Chao Chen","hidden":false},{"_id":"6a0fc10ea53a61ce2e422c9d","name":"Longfei Xu","hidden":false},{"_id":"6a0fc10ea53a61ce2e422c9e","name":"Kaikui Liu","hidden":false},{"_id":"6a0fc10ea53a61ce2e422c9f","name":"Xiangxiang Chu","hidden":false}],"publishedAt":"2026-05-21T00:00:00.000Z","submittedOnDailyAt":"2026-05-22T00:00:00.000Z","title":"TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation","submittedOnDailyBy":{"_id":"66d255e3947594430c723ff6","avatarUrl":"/avatars/c56e4792332a01bf34085a75ee64916e.svg","isPro":false,"fullname":"xiaochonglinghu","user":"xiaochonglinghu","type":"user","name":"xiaochonglinghu"},"summary":"Public transit route planning traditionally depends on structured map infrastructure and complex routing engines, and no existing dataset supports training models to bypass this dependency. We present TransitLM, a large-scale dataset of over 13 million transit route planning records from four Chinese cities covering 120,845 stations and 13,666 lines, released as a continual pre-training corpus and benchmark data for three evaluation tasks with complementary metrics. Experiments show that an LLM trained on TransitLM produces structurally valid routes at high accuracy and implicitly grounds arbitrary GPS coordinates to appropriate stations without any explicit mapping. These results demonstrate that transit route planning can be learned entirely from data, enabling end-to-end, map-free route generation directly from origin-destination information. The dataset and benchmark are available at https://huggingface.co/datasets/GD-ML/TransitLM, with evaluation code at https://github.com/HotTricker/TransitLM.","upvotes":105,"discussionId":"6a0fc10ea53a61ce2e422ca0","githubRepo":"https://github.com/HotTricker/TransitLM","githubRepoAddedBy":"user","ai_summary":"TransitLM dataset enables end-to-end transit route planning using large language models trained on structured transit data, eliminating the need for traditional map-based approaches.","ai_keywords":["large language models","transit route planning","structured map infrastructure","routing engines","continual pre-training","evaluation tasks","GPS coordinates","origin-destination information"],"githubStars":96,"organization":{"_id":"67d11771890254196d3174e5","name":"GD-ML","fullname":"AMAP-ML","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/67d116c47be76de1a40873ca/s5ukAx9E36ZZIKvbpBRi4.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"682eeb11d28d9650c926abb0","avatarUrl":"/avatars/3c8a34cc61aa2330ffcb52446a9afe70.svg","isPro":false,"fullname":"LONGFEI XU","user":"Xufew","type":"user"},{"_id":"6964a5cb0be2016837983f50","avatarUrl":"/avatars/c99715fcb6c273184416bdb9852358cf.svg","isPro":false,"fullname":"DuanYihai","user":"Shenhaihahaha","type":"user"},{"_id":"6964a38ae811a5a53de8ebc9","avatarUrl":"/avatars/d2d3ed0ce3e25a158c6863a15ebeb25b.svg","isPro":false,"fullname":"So","user":"dwdSo","type":"user"},{"_id":"68ca4e012b37a7b211ff0c23","avatarUrl":"/avatars/8c7f6a2efb7d169cfe409db489963739.svg","isPro":false,"fullname":"AiSniper","user":"AiSniper2025","type":"user"},{"_id":"6964c5ef0342c2fce9136a9f","avatarUrl":"/avatars/ea7fa3f2d4b8045132675c7f1c60b5ce.svg","isPro":false,"fullname":"Chengran","user":"chengran0115","type":"user"},{"_id":"689ae74c60236ebade8018bc","avatarUrl":"/avatars/6ee947dabaf12181518595e6193ff6ae.svg","isPro":false,"fullname":"Huimin Yan","user":"Hannah6huimin","type":"user"},{"_id":"6964a33d2ab4d967d0543d84","avatarUrl":"/avatars/93702bd3303bccb4ee29812e3025e7aa.svg","isPro":false,"fullname":"liuzheng","user":"lzlz2000","type":"user"},{"_id":"6964ab352fb9f77cbe8c9854","avatarUrl":"/avatars/2c4734b40bb98d162837fbe6b128f7d7.svg","isPro":false,"fullname":"Hanyu Guo","user":"MaoPaJia","type":"user"},{"_id":"695c72dbd0638f21b7f7c645","avatarUrl":"/avatars/2c7d50b623859d4f47f3a6ac20145982.svg","isPro":false,"fullname":"sunjunjie","user":"j2sun","type":"user"},{"_id":"6964ac022fb9f77cbe8cac95","avatarUrl":"/avatars/adbe319c1a1b85c09ac3854b82ca1bfc.svg","isPro":false,"fullname":"luo","user":"weiluo2068","type":"user"},{"_id":"69649c726a20f62862cf9578","avatarUrl":"/avatars/b5314840d74265d3fa4f88948146a729.svg","isPro":false,"fullname":"liukaikui","user":"liukaikui","type":"user"},{"_id":"6448f74f011671fb7bdbf5b7","avatarUrl":"/avatars/b9c883ab3966889386d89f7db75ef1af.svg","isPro":false,"fullname":"ScottZhang","user":"ScottZhang","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"67d11771890254196d3174e5","name":"GD-ML","fullname":"AMAP-ML","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/67d116c47be76de1a40873ca/s5ukAx9E36ZZIKvbpBRi4.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.22355.md"}">

Papers

arxiv:2605.22355

TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation

Published on May 21

· Submitted by

xiaochonglinghu on May 22

AMAP-ML

Upvote

105

Authors:

Abstract

TransitLM dataset enables end-to-end transit route planning using large language models trained on structured transit data, eliminating the need for traditional map-based approaches.

AI-generated summary

Public transit route planning traditionally depends on structured map infrastructure and complex routing engines, and no existing dataset supports training models to bypass this dependency. We present TransitLM, a large-scale dataset of over 13 million transit route planning records from four Chinese cities covering 120,845 stations and 13,666 lines, released as a continual pre-training corpus and benchmark data for three evaluation tasks with complementary metrics. Experiments show that an LLM trained on TransitLM produces structurally valid routes at high accuracy and implicitly grounds arbitrary GPS coordinates to appropriate stations without any explicit mapping. These results demonstrate that transit route planning can be learned entirely from data, enabling end-to-end, map-free route generation directly from origin-destination information. The dataset and benchmark are available at https://huggingface.co/datasets/GD-ML/TransitLM, with evaluation code at https://github.com/HotTricker/TransitLM.