PhysDesignBench
GitHub
Reimplementation task · v4-canonical-j-simulated-envelope

Dual-Trident SWG-v0

Dual-Trident SWG Edge Coupler Implementation/Optimization

Implement and optimize a silicon dual-trident subwavelength-grating edge coupler over 1536–1628 nm. The trusted evaluator measures Gaussian-beam coupling into the output waveguide's fundamental TE mode.

Runs
11
Scored finals
11 of 11
Best final
S = 0.37 Opus 5.5 · Claude Code
Score over timeBest-so-far score S for each run · small dots are probes, large dots final submissions
Score S from 0 to 1 against elapsed minutes. Final scores: Opus 5.5 (Claude Code) 0.366; Fable 5.1 (Claude Code) 0.359; GPT-6.1 Sol (Codex) 0.353; Gemini 4 Argon (mini-SWE-agent) 0.348; GPT-6 Astra (Codex) 0.346; Sonnet 5.5 (high) (Claude Code) 0.345; Kimi K3.1 Pro (Claude Code) 0.318; Muse Wonder (high) (Muse Code) 0.278; GLM 5.3 Flash (max) (mini-SWE-agent) 0.179; GPT-5.6 Sol (Codex) 0.140; GLM 5.3 Flash (high) (Claude Code) 0.109.0.000.250.500.751.000153045607590105120135Elapsed time (min)Score SGLM 5.3 Flash (high) · Claude Code: final S = 0.109 at 34 minGPT-5.6 Sol · Codex: final S = 0.140 at 120 minGLM 5.3 Flash (max) · mini-SWE-agent: final S = 0.179 at 100 minMuse Wonder (high) · Muse Code: final S = 0.278 at 43 minKimi K3.1 Pro · Claude Code: final S = 0.318 at 121 minSonnet 5.5 (high) · Claude Code: final S = 0.345 at 40 minGPT-6 Astra · Codex: final S = 0.346 at 120 minGemini 4 Argon · mini-SWE-agent: final S = 0.348 at 121 minGPT-6.1 Sol · Codex: final S = 0.353 at 120 minFable 5.1 · Claude Code: final S = 0.359 at 112 minOpus 5.5 · Claude Code: final S = 0.366 at 95 min

Showing 11 of 11 runs. Hover a line or a model to isolate it; click a model to hide or show it.

Final layouts

Every run’s final GDS, at one scale.

The design each run ended with, drawn from its evaluated GDS, with the task’s ports, modes and band as the evaluator defines them, the run’s own length, and the published result that S = 0.5 stands for. Every layout uses the same physical scale, so sizes compare directly. Hatched layouts failed a validity gate and score zero. Select a layout to zoom and toggle its layers.

  • Optical mask

  • Dual-Trident SWG-v0 · Claude Code · Claude Opus 5.5

    S = 0.366 · scale bar 10 µm

    Fit
    Opus 5.5Claude CodeS = 0.37

  • Dual-Trident SWG-v0 · Claude Code · Claude Fable 5.1

    S = 0.359 · scale bar 10 µm

    Fit
    Fable 5.1Claude CodeS = 0.36

  • Dual-Trident SWG-v0 · Codex · GPT-6.1-sol

    S = 0.353 · scale bar 10 µm

    Fit
    GPT-6.1 SolCodexS = 0.35

  • Dual-Trident SWG-v0 · mini-SWE-agent · Gemini-4-Argon

    S = 0.348 · scale bar 10 µm

    Fit
    Gemini 4 Argonmini-SWE-agentS = 0.35

  • Dual-Trident SWG-v0 · Codex · GPT-6 Astra

    S = 0.346 · scale bar 10 µm

    Fit
    GPT-6 AstraCodexS = 0.35

  • Dual-Trident SWG-v0 · Claude Code · Claude Sonnet 5.5

    S = 0.345 · scale bar 10 µm

    Fit
    Sonnet 5.5 (high)Claude CodeS = 0.35

  • Dual-Trident SWG-v0 · Claude Code · Kimi K3.1 Pro

    S = 0.318 · scale bar 10 µm

    Fit
    Kimi K3.1 ProClaude CodeS = 0.32

  • Dual-Trident SWG-v0 · Muse Code · Muse Wonder

    S = 0.278 · scale bar 10 µm

    Fit
    Muse Wonder (high)Muse CodeS = 0.28

  • Dual-Trident SWG-v0 · mini-SWE-agent · GLM 5.3 Flash

    S = 0.179 · scale bar 10 µm

    Fit
    GLM 5.3 Flash (max)mini-SWE-agentS = 0.18

  • Dual-Trident SWG-v0 · Codex · GPT-5.6-sol

    S = 0.140 · scale bar 10 µm

    Fit
    GPT-5.6 SolCodexS = 0.14

  • Dual-Trident SWG-v0 · Claude Code · GLM 5.3 Flash

    S = 0.109 · scale bar 10 µm

    Fit
    GLM 5.3 Flash (high)Claude CodeS = 0.11