GD-ML/TransitLM

1---2license: cc-by-nc-4.03task_categories:4- text-generation5language:6- zh7tags:8- transportation9- route-planning10- public-transit11- mobility12- instruction-tuning13- benchmark14pretty_name: TransitLM15---16 17# TransitLM: Dataset Release & Evaluation Protocol18 19## Dataset Description20 21TransitLM is a dataset for public transit route planning in Chinese urban environments, designed to support training and evaluation of language models that generate structured transit routes from origin-destination information. The full dataset covers four cities: Beijing, Shanghai, Shenzhen, and Chengdu, and includes coordinates, station sequences, transfer structure, line information, and route annotations.22 23### Links24 25- Demo: [https://transit-lm.amap.com/](https://transit-lm.amap.com/)26- GitHub: [https://github.com/HotTricker/TransitLM](https://github.com/HotTricker/TransitLM)27- arXiv: [https://arxiv.org/abs/2605.22355](https://arxiv.org/abs/2605.22355)28- Hugging Face Paper: [https://huggingface.co/papers/2605.22355](https://huggingface.co/papers/2605.22355)29- ModelScope: [https://modelscope.cn/datasets/GD-ML/TransitLM](https://modelscope.cn/datasets/GD-ML/TransitLM)30 31### Released Files32 33| File Pattern | Content | Section |34|---|---|---|35| `transit-lm-benchmark*-train.csv` | Anonymized SFT training splits (benchmark-aligned) | §2 |36| `transit-lm-benchmark*-test.csv` | Anonymized SFT test splits for task evaluation | §2 |37| `transit-lm-general-benchmark*-test.csv` | Real-world benchmark test sets (real station names + coordinates) | §1 |38| `transit-lm-station-info.csv` | Static station-level reference information | — |39| `transit-lm-station-ids.json` | Station ID list covering the released transit network | — |40| `transit-lm-cpt-corpus/transit-lm-cpt-corpus-part*.csv` | CPT corpus shards with unified three-digit sequential filenames (e.g. `part001`) | §2 |41 42All released data follows the anonymization principles described in our paper. The sections below explain the data treatment and evaluation protocol for each component.43 44### File Format Examples45 46#### 1. Anonymized SFT train/test (`transit-lm-benchmark*-train.csv`, `transit-lm-benchmark*-test.csv`)47 48Columns:49`task_id, req_type, adcode, indexid, system_prompt, sft_prompt, sft_label`50 51Example:52 53```csv54task_id,req_type,adcode,indexid,system_prompt,sft_prompt,sft_label55benchmark1-train,0,310110,a02385072c754b19067c70af8bb62de1,"你是公共交通路线规划专家，熟悉公共交通网络。...","{""query"":""我需要从上海北外滩亚朵酒店到上海市公安局治安总队一大队，请告诉我乘车方案"",""start"":""121.515027,31.260541"",""end"":""121.490488,31.209797"",""city"":""上海""}","{""line_sequence"":[""地铁4号线内圈(宜山路--宜山路)""],""station_sequence"":[""bus-1669599"",""bus-1670350"",""bus-1670805""],""total_distance"":""10.7公里"",""total_time"":""44分钟"",""total_fare"":""4"",""start_transfer_mode"":""步行"",""end_transfer_mode"":""步行"",""start_transfer_distance"":""523米"",""end_transfer_distance"":""1.1公里""}"56```57 58#### 2. Real-world benchmark (`transit-lm-general-benchmark*-test.csv`)59 60Columns:61`task_id, req_type, adcode, indexid, system_prompt, sft_prompt, sft_label`62 63Example:64 65```csv66task_id,req_type,adcode,indexid,system_prompt,sft_prompt,sft_label67general_benchmark1-test,0,440307,04258df410d3ec5d53e64de9f49b26be,"你是公共交通路线规划专家，熟悉公共交通网络。...","{""query"":""我在佳兆业君汇公馆，想去石芽岭(地铁站)，怎么坐车？"",""start"":""114.130533,22.692349"",""end"":""114.137989,22.628479"",""city"":""深圳""}","{""line_sequence"":[""977路(启英学校--田贝总站)""],""station_sequence"":[""平湖派出所"",""平湖中心小学"",""富伟厂""],""total_distance"":""10.3公里"",""total_time"":""46分钟"",""total_fare"":""2"",""start_transfer_mode"":""步行"",""end_transfer_mode"":""步行"",""start_transfer_distance"":""185米"",""end_transfer_distance"":""77米""}"68```69 70#### 3. Station reference (`transit-lm-station-info.csv`)71 72Columns:73`stop_id, ad_code, coord_x, coord_y, next_hop_stations`74 75Example:76 77```csv78stop_id,ad_code,coord_x,coord_y,next_hop_stations79bus-1000071,110117,117.090101,40.130034,"[""bus-1000202"", ""bus-1000089"", ""bus-1000087"", ""bus-1000071""]"80```81 82#### 4. Station ID list (`transit-lm-station-ids.json`)83 84Example:85 86```json87[88  "bus-1000071",89  "bus-1000074",90  "bus-1000082"91]92```93 94#### 5. CPT corpus shard (`transit-lm-cpt-corpus/transit-lm-cpt-corpus-part*.csv`)95 96Columns:97`id, content, pt`98 99The `pt` field is used only as a partition marker inherited from the original sharding pipeline. It carries no semantic meaning for modeling or evaluation.100 101Example:102 103```csv104id,content,pt105session_65703461_0,"路线规划。规划城市：北京，起点坐标：116.402471,39.933658，终点坐标：116.068893,39.648444。有以下几种方案：方案1（纯打车）... 方案2（公交+地铁）... 经过bus-900579，bus-898547，...，bus-857245",p01106```107 108**Overview.** We release two parallel test sets that correspond one-to-one, serving different model categories:109 110- **Anonymized SFT test set** (30K samples). For models trained on our anonymized corpus. These models have only seen numeric station IDs and perturbed coordinates, so they must be evaluated in the same representation space.111- **Real-world benchmark** (30K samples). For general-purpose LLMs not trained on our data. These models operate with real station names and coordinates and cannot interpret anonymized IDs, so they require unperturbed evaluation data.112 113This parallel design ensures fair evaluation for both model types while keeping results directly comparable.114 115---116 117## 1. Benchmark Test Samples118 11930,000 test examples across three tasks: optimal route generation, preference-aware planning, and multi-route generation.120 121**Data characteristics:**122- Real GPS coordinates without perturbation123- Real station names in natural language124- Full route-structural metadata: station sequences, line sequences, transfer points, distances, travel times, and fares125 126**Purpose.** This subset is the ground-truth evaluation benchmark for the research community. The CPT/SFT training corpus applies coordinate perturbation, so evaluating general-purpose LLMs on perturbed data would not reflect real-world performance. Unperturbed samples enable fair comparison across models trained on our corpus, tool-augmented LLMs, and any future transit-domain model.127 128**Privacy.** These samples contain no user identifiers, no timestamps, and no cross-record linkage. Each is an isolated OD request. All station coordinates and names are already publicly accessible through mapping APIs, so releasing them in benchmark format introduces no additional privacy risk.129 130---131 132## 2. CPT + SFT Corpus (Anonymized)133 13413.9M CPT records, 90K SFT training examples, and 30K SFT test examples.135 136### Anonymization137 138| Transformation | Method | Training Impact |139|---|---|---|140| Coordinate perturbation | Deterministic hash-based offset, ±100m on a ~500m grid | Negligible. Grid-coherent offset preserves spatial structure. Error < 0.1% of mean route length |141| Station name removal | Names removed from station sequences, retaining only numeric IDs | None. Topological learning relies on IDs and connectivity |142 143**Rationale.** Bulk harvesting coordinate-ID mappings from a 13.9M-record corpus could enable unauthorized commercial use. OD coordinates reflect real user trip intentions. Perturbation combined with structural safeguards prevents re-identification.144 145### SFT Test Set146 147The 30K SFT test samples undergo identical anonymization. See the Overview above for how they relate to the real-world benchmark and which model category uses which test set.148 149### Usability150 151Despite anonymization, this corpus fully supports continual pre-training for transit topology and spatial reasoning, supervised fine-tuning for all three benchmark tasks, and reproducibility of all paper results.152 153---154 155## 3. Privacy Architecture156 157Unlike GPS trajectory datasets such as T-Drive or GeoLife that record continuous multi-day traces enabling re-identification, each record here is an isolated OD planning request with no temporal continuity.158 159Structural safeguards enforced at the data construction stage:160- Single calendar day sampling with no timestamps retained161- User identifiers removed prior to construction, with no linkage key across records162- Only route-structural metadata released. No demographic attributes, device fingerprints, or PII163 164These properties make it infeasible to associate records with the same individual or reconstruct mobility patterns, regardless of coordinate perturbation.165 166---167 168## 4. Why Both Perturbed and Real Data?169 170| Concern | Anonymized corpus | Benchmark samples |171|---|---|---|172| Coordinate privacy | ✅ ±100m perturbation prevents bulk harvesting | Public data, no new risk |173| Fair evaluation | ❌ Perturbed coordinates disadvantage general LLMs | ✅ Real coordinates enable direct comparison |174| Reproducibility | ✅ Full training pipeline reproducible | ✅ Standardized community test set |175| Training utility | ✅ < 0.1% spatial error, topology preserved | Not used for training |176 177This dual-release design maximizes research utility through reproducible training and fair evaluation, while minimizing privacy exposure by applying perturbation only where data scale creates risk.178