PhysDesignBench
GitHub
Reimplementation task · v1-binary-paper-fom-j

HMDM polarization splitter–rotator

HMDM Polarization Splitter–Rotator Implementation/Optimization

Optimize the paper-derived polarization splitter–rotator's independently selectable nanohole sites. The trusted evaluator measures desired-mode transmission for TE₀ and TM₀ inputs across 1500–1600 nm.

Runs
11
Scored finals
11 of 11
Best final
S = 0.12 GPT-6.1 Sol · Codex
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: GPT-6.1 Sol (Codex) 0.118; GPT-6 Astra (Codex) 0.108; Sonnet 5.5 (high) (Claude Code) 0.104; GPT-5.6 Sol (Codex) 0.098; Opus 5.5 (Claude Code) 0.094; Kimi K3.1 Pro (Claude Code) 0.094; GLM 5.3 Flash (high) (Claude Code) 0.093; Muse Wonder (high) (Muse Code) 0.092; Fable 5.1 (Claude Code) 0.092; Gemini 4 Argon (mini-SWE-agent) 0.091; GLM 5.3 Flash (max) (mini-SWE-agent) 0.091.0.000.250.500.751.000153045607590105120135Elapsed time (min)Score SGemini 4 Argon · mini-SWE-agent: final S = 0.091 at 121 minGLM 5.3 Flash (max) · mini-SWE-agent: final S = 0.091 at 97 minFable 5.1 · Claude Code: final S = 0.092 at 61 minMuse Wonder (high) · Muse Code: final S = 0.092 at 42 minGLM 5.3 Flash (high) · Claude Code: final S = 0.093 at 81 minKimi K3.1 Pro · Claude Code: final S = 0.094 at 121 minOpus 5.5 · Claude Code: final S = 0.094 at 99 minGPT-5.6 Sol · Codex: final S = 0.098 at 121 minSonnet 5.5 (high) · Claude Code: final S = 0.104 at 73 minGPT-6 Astra · Codex: final S = 0.108 at 121 minGPT-6.1 Sol · Codex: final S = 0.118 at 122 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

  • HMDM polarization splitter–rotator · Codex · GPT-6.1-sol

    S = 0.118 · scale bar 2 µm

    Fit
    GPT-6.1 SolCodexS = 0.12

  • HMDM polarization splitter–rotator · Codex · GPT-6 Astra

    S = 0.108 · scale bar 2 µm

    Fit
    GPT-6 AstraCodexS = 0.11

  • HMDM polarization splitter–rotator · Claude Code · Claude Sonnet 5.5

    S = 0.104 · scale bar 2 µm

    Fit
    Sonnet 5.5 (high)Claude CodeS = 0.10

  • HMDM polarization splitter–rotator · Codex · GPT-5.6-sol

    S = 0.098 · scale bar 2 µm

    Fit
    GPT-5.6 SolCodexS = 0.10

  • HMDM polarization splitter–rotator · Claude Code · Claude Opus 5.5

    S = 0.094 · scale bar 2 µm

    Fit
    Opus 5.5Claude CodeS = 0.09

  • HMDM polarization splitter–rotator · Claude Code · Kimi K3.1 Pro

    S = 0.094 · scale bar 2 µm

    Fit
    Kimi K3.1 ProClaude CodeS = 0.09

  • HMDM polarization splitter–rotator · Claude Code · GLM 5.3 Flash

    S = 0.093 · scale bar 2 µm

    Fit
    GLM 5.3 Flash (high)Claude CodeS = 0.09

  • HMDM polarization splitter–rotator · Muse Code · Muse Wonder

    S = 0.092 · scale bar 2 µm

    Fit
    Muse Wonder (high)Muse CodeS = 0.09

  • HMDM polarization splitter–rotator · Claude Code · Claude Fable 5.1

    S = 0.092 · scale bar 2 µm

    Fit
    Fable 5.1Claude CodeS = 0.09

  • HMDM polarization splitter–rotator · mini-SWE-agent · Gemini-4-Argon

    S = 0.091 · scale bar 2 µm

    Fit
    Gemini 4 Argonmini-SWE-agentS = 0.09

  • HMDM polarization splitter–rotator · mini-SWE-agent · GLM 5.3 Flash

    S = 0.091 · scale bar 2 µm

    Fit
    GLM 5.3 Flash (max)mini-SWE-agentS = 0.09