Problem
Research questions and friction points this paper is trying to address.
road safety
uncoded crash factors
police crash records
narrative analysis
traffic safety estimation
Estimating Uncoded Crash Factors with Tabular Foundation and System One Models: Kumo Tabular and Jev
🤖 AI Summary
This study addresses the critical information loss in coded fields of traffic crash records, where narrative text details are frequently omitted, resulting in incomplete safety statistics and unquantifiable biases. To overcome this, it proposes a "dual-perspective calibrated reader" framework that integrates a tabular foundation model (Kumo Tabular) with a narrative parsing model (Jev). By combining stratified probability sampling, multi-wave prediction debiasing estimation, and human calibration, the approach enables efficient synergy between automated models and expert judgment. The proposed method successfully identifies high-injury disparities associated with uncoded factors such as mobile phone use, reducing the half-width of estimation error to 16.2%. Furthermore, it achieves a fifteen-fold improvement in processing speed over TabPFN, substantially lowering the costs of full-scale manual review while providing an empirically validated rereading checklist.
📝 Abstract
Road safety programs count the coded fields of police crash records, while the officer's narrative, which often records factors the fields omit, is rarely read. A safety office thus cannot tell how much its counts miss or where to review. This study develops and evaluates a system that joins both views of the 5,601,890 Texas crashes from 2017 to 2025 into population estimates with stated validity. An in-context tabular foundation model, Kumo Tabular, reads the coded record of every crash, a calibrated System One model, Jev, reads the narratives of two probability samples, and human judgments recalibrate its probabilities. A multiwave predict-then-debias estimator joins the three tiers, and a second human tier drawn with recorded probabilities checks the estimates by design. For hydroplaning, medical episodes, fatigue, animals, and phone use, the narrative documents more injury crashes than the coded field, 15,074 against 7,340 for phone use, and the human check agrees with all fifteen estimates within its margin. A re-read list ranked by Kumo Tabular finds confirmed discordance 7 to 58 times as often as random reading. At the planning cost of human coding, one further round of human judgments would cut the root mean square relative half-width from 22.0 to 16.2 percent, against 21.2 for reading every narrative. Two calibrated readers of different views, joined by a sampling design, give a safety office counts, a discordance map, a validated re-read list, and a reading budget, with Kumo Tabular reading the table at 15 times the speed of TabPFN 3.5.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Tabular Foundation Model
Predict-then-debias Estimator
System One Model
In-context Learning
Uncoded Crash Factors
🔎 Similar Papers
No similar papers found.
