mlsanityData to 7 Oct 2026

ALCo-FM: Adaptive Long-Context Foundation Model for Accident Prediction

arXiv 2507.08153 ↗First posted 10 Jul 2025cs.LG cs.AIcited by 1

On the map: LLM reasoning → Chain-of-thought reasoning for driving

Abstract (arXiv)

Traffic accidents are rare, yet high-impact events that require long-context multimodal reasoning for accurate risk forecasting. In this paper, we introduce ALCo-FM, a unified adaptive long-context foundation model that computes a volatility pre-score to dynamically select context windows for input data and encodes and fuses these multimodal data via shallow cross attention. Following a local GAT layer and a BigBird-style sparse global transformer over H3 hexagonal grids, coupled with Monte Carlo dropout for confidence, the model yields superior, well-calibrated predictions. Trained on data from 15 US cities with a class-weighted loss to counter label imbalance, and fine-tuned with minimal data on held-out cities, ALCo-FM achieves 0.94 accuracy, 0.92 F1, and an ECE of 0.04, outperforming more than 20 state-of-the-art baselines in large-scale urban risk prediction. Code and dataset are available at: https://github.com/PinakiPrasad12/ALCo-FM

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Chain-of-thought reasoning for driving, one of 17 directions in LLM reasoning. Builds on 26 of its 60 references that are in the map. First posted on arXiv 10 Jul 2025.

Builds on

26 of 60 references.

References · 26

  1. 2016XGBoost: A Scalable Tree Boosting Systemcited by 5,120
  2. 2017Billion-scale similarity search with GPUscited by 2,477
  3. 2017Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecastingcited by 1,125
  4. 2021Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNetcited by 1,096
  5. 2019Graph WaveNet for Deep Spatial-Temporal Graph Modelingcited by 820
  6. 2016Adaptive Computation Time for Recurrent Neural Networkscited by 479
  7. 2020Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecastingcited by 212
  8. 2021Spatial-Temporal Graph ODE Networks for Traffic Flow Forecastingcited by 209
  9. 2022SAITS: Self-Attention-based Imputation for Time Seriescited by 167
  10. 2021Dynamic Graph Convolutional Recurrent Network for Traffic Prediction: Benchmark and Solutioncited by 156
  11. 2019Adaptive Attention Span in Transformerscited by 148
  12. 2020Uncertainty-based Traffic Accident Anticipation with Spatio-Temporal Relational Learningcited by 91
14 more
  1. 2024ViT-CoMer: Vision Transformer with Convolutional Multi-scale Feature Interaction for Dense Predictionscited by 65
  2. 2024ContiFormer: Continuous-Time Transformer for Irregular Time Series Modelingcited by 61
  3. 2023Fully-Connected Spatial-Temporal Graph for Multivariate Time-Series Datacited by 23
  4. 2023MST-GAT: A Multimodal Spatial-Temporal Graph Attention Network for Time Series Anomaly Detectioncited by 20
  5. 2019A Countrywide Traffic Accident Datasetcited by 18
  6. 2022Automated Dilated Spatio-Temporal Synchronous Graph Modeling for Traffic Predictioncited by 15
  7. 2024GTC: GNN-Transformer Co-contrastive Learning for Self-supervised Heterogeneous Graph Representationcited by 9
  8. 2024Urban Traffic Accident Risk Prediction Revisited: Regionality, Proximity, Similarity and Sparsitycited by 6
  9. 2024CrashFormer: A Multimodal Architecture to Predict the Risk of Crashcited by 5
  10. 2021Risk Prediction on Traffic Accidents using a Compact Neural Model for Multimodal Information Fusion over Urban Big Datacited by 3
  11. 2022Will there be a construction? Predicting road constructions based on heterogeneous spatiotemporal datacited by 3
  12. 2023Iterative missing value imputation based on feature importancecited by 2
  13. 2024SST-GCN: The Sequential based Spatio-Temporal Graph Convolutional networks for Minute-level and Road-level Traffic Accident Risk Predictioncited by 2
  14. 2024DocParseNet: Advanced Semantic Segmentation and OCR Embeddings for Efficient Scanned Document Annotationcited by 2

Built on by

Cited by year

Citing papers per year: 2025: 1.

1 paper.

  1. 2025DocVAL: Validated Chain-of-Thought Distillation for Grounded Document VQAcited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2024Learning Traffic Crashes as Language: Datasets, Benchmarks, and What-if Causal Analyses
  2. 2026Forecasting the Emergence and Evolution of Crash Hotspots: A Unified Deep Learning Framework for Proactive Traffic Safety
  3. 2026MambaLSTM: A Spatio-Temporal Framework for Enhanced Traffic Accident Risk Prediction
  4. 2026TransportAgents: a multi-agents LLM framework for traffic accident severity prediction
  5. 2025Multimodal Large Language Models for Enhanced Traffic Safety: A Comprehensive Review and Future Trends
  6. 2025Large Foundation Models for Trajectory Prediction in Autonomous Driving: A Comprehensive Survey
  7. 2025MARBLE: A Multi-Agent Rule-Based LLM Reasoning Engine for Accident Severity Prediction
  8. 2025From Street Views to Urban Science: Discovering Road Safety Factors with Multimodal Large Language Models