2d-fet-bench: from spatial reasoning to fet design on flakes
Abstract
Field-effect transistor (FET) layouts on exfoliated two-dimensional flakes are typically drawn by hand for each flake, placing contacts and gates to match its position and outline in optical micrographs. To our knowledge, no executable benchmark tests whether language-model agents can perform this flake-specific construction reliably. We introduce 2D-FET-Bench V2, a benchmark of 128 layout tasks built from microscopy-derived flake contours, including hole-containing flakes and multi-flake tasks. Each task supplies a textual device specification and contour coordinates. An agent generates typed polygon and path operations rendered to GDSII. A deterministic verifier checks geometric and structural requirements, and a separate integrity check verifies that the supplied contours remain unchanged. Scripted reference layouts pass all 128 tasks, showing that every task is solvable. We evaluate six models and seven workflow and scaffold variants of GPT5.6-Luna, with five attempts per task. The best-performing configuration in the six-model panel, GPT5.6-Luna with ReAct-3, passes 62.3% of attempts and solves 80.5% of tasks at least once (coverage) and 43.8% in all five attempts (consistency). ReAct-3 exceeds the one-pass Plan-and-Execute by 27.0 pass@1 points at 2.46 times the tokens. An expert audit of one sampled verifier-passing layout per covered task, across five ReAct-3 configurations, accepts 56.4% to 63.5% of them. The benchmark evaluates geometric and structural FET layout construction.
Related
- Scaling LLM Agents for Materials Design through Hierarchical Collective Reasoning
- Optical design and characterization of geometric distortion and spatial resolution in lens-enhanced aerial imaging by retro-reflection (LeAIRR)
- Interaction-driven dynamics in graphene flakes as a benchmark for quantum simulation
- Chiral symmetry and the orbital susceptibility of disordered honeycomb flakes
- Evidence for an effective gauge field in flakes of superconducting 4Hb-TaS2
BibTeX
@misc{zhou20262dfetbench,
title = {{2d-fet-bench: from spatial reasoning to fet design on flakes}},
author = {Dunzhi Zhou and Chengyu Zhu and Gang Qiu and Caiwen Ding},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.07423},
eprint = {2610.07423},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.07423},
}
RIS
TY - UNPB
TI - 2d-fet-bench: from spatial reasoning to fet design on flakes
AU - Dunzhi Zhou
AU - Chengyu Zhu
AU - Gang Qiu
AU - Caiwen Ding
PY - 2026
DA - 2026/10/05
PB - arXiv
DO - 10.48550/arxiv.2610.07423
UR - https://arxiv.org/abs/2610.07423
AB - Field-effect transistor (FET) layouts on exfoliated two-dimensional flakes are typically drawn by hand for each flake, placing contacts and gates to match its position and outline in optical micrographs. To our knowledge, no executable benchmark tests whether language-model agents can perform this flake-specific construction reliably. We introduce 2D-FET-Bench V2, a benchmark of 128 layout tasks built from microscopy-derived flake contours, including hole-containing flakes and multi-flake tasks. Each task supplies a textual device specification and contour coordinates. An agent generates typed polygon and path operations rendered to GDSII. A deterministic verifier checks geometric and structural requirements, and a separate integrity check verifies that the supplied contours remain unchanged. Scripted reference layouts pass all 128 tasks, showing that every task is solvable. We evaluate six models and seven workflow and scaffold variants of GPT5.6-Luna, with five attempts per task. The best-performing configuration in the six-model panel, GPT5.6-Luna with ReAct-3, passes 62.3% of attempts and solves 80.5% of tasks at least once (coverage) and 43.8% in all five attempts (consistency). ReAct-3 exceeds the one-pass Plan-and-Execute by 27.0 pass@1 points at 2.46 times the tokens. An expert audit of one sampled verifier-passing layout per covered task, across five ReAct-3 configurations, accepts 56.4% to 63.5% of them. The benchmark evaluates geometric and structural FET layout construction.
ER -