The Year in AI Papers
The year/Independent research

Paper 2604.03144

InCoder-32B-Thinking: Industrial Code World Model for Thinking

Jian Yang, Wei Zhang, Jiajun Wu, Junhang Cheng, Tuney Zheng, Fanglin Xu, Weicheng Gu, Lin Jing, Yaxin Du, Joseph Li, Yizhi Li, Yan Xing, et al.

Published
Apr 2026
Research lab
Independent
Citations
3
GitHub
Not linked

01 In brief

Summary

The paper introduces InCoder-32B-Thinking, a 32B-parameter code model trained to reason about industrial software development (chip design, GPU optimization, embedded systems, 3D modeling) by integrating an Error-driven Chain-of-Thought (ECoT) synthesis framework and an Industrial Code World Model (ICWM).

ECoT generates reasoning traces by modeling error-correction processes from multi-turn dialogues with real execution feedback, while ICWM learns causal dynamics between code and hardware behavior from domain-specific execution traces (e.g., Verilog simulation, GPU profiling), enabling self-verification and large-scale synthetic data generation without repeated toolchain execution.

The model is trained on 540M tokens of thinking data, combining real and ICWM-simulated trajectories.

Evaluations on 14 general and 9 industrial benchmarks show top-tier open-source results: 81.3% on LiveCodeBench V5, 84.0% on CAD-Coder, 38.0% on KernelBench L2, and 70.4% on SWE-bench Verified.

ICWM achieves 96.7% outcome prediction accuracy and 94.4% trajectory agreement across domains.

Scaling thinking data from 180M to 540M tokens improves industrial metrics, e.g., VeriScope score rises from 61.8 to 75.4.

The model adaptively calibrates thinking depth, with median thinking lengths ranging from 91 characters (agentic coding) to 19,015 characters (GPU optimization).

02 From the paper

Abstract

Industrial software development across chip design, GPU optimization, and embedded systems lacks expert reasoning traces showing how engineers reason about hardware constraints and timing semantics. In this work, we propose InCoder-32B-Thinking, trained on the data from the Error-driven Chain-of-Thought (ECoT) synthesis framework with an industrial code world model (ICWM) to generate reasoning traces. Specifically, ECoT generates reasoning chains by synthesizing the thinking content from multi-turn dialogue with environmental error feedback, explicitly modeling the error-correction process. ICWM is trained on domain-specific execution traces from Verilog simulation, GPU profiling, etc., learns the causal dynamics of how code affects hardware behavior, and enables self-verification by predicting execution outcomes before actual compilation. All synthesized reasoning traces are validated through domain toolchains, creating training data matching the natural reasoning depth distribution of industrial tasks. Evaluation on 14 general (81.3% on LiveCodeBench v5) and 9 industrial benchmarks (84.0% in CAD-Coder and 38.0% on KernelBench) shows InCoder-32B-Thinking achieves top-tier open-source results across all domains.GPU Optimization