PhysDesignBench
GitHub
RF preview task · HFSS · diagnostic scores, not in the paper's 26 tasks

Dual-band Filter-v0

Dual-band waveguide filter design

Design a compact, all-metal waveguide filter with two 150 MHz service bands at 11.925–12.075 and 13.925–14.075 GHz. Rejection is evaluated over 10.5–11.3, 12.5–13.5 and 14.5–15 GHz. Trusted Ansys HFSS simulations measure both bands and the shaped enclosure volume.

Runs
2
Scored finals
2 of 2
Best final
S = 0.62 GPT-6.1 Sol · Codex
Score over timeBest-so-far diagnostic score S for each run · small dots are probes, large dots final submissions
Diagnostic score S from 0 to 1 against elapsed minutes. Final scores: GPT-6.1 Sol (Codex) 0.620; Opus 5.5 (Claude Code) 0.568.0.000.250.500.751.000306090120150180210240270300330360390420450480Elapsed time (min)Diagnostic score SOpus 5.5 · Claude Code: final S = 0.568 at 468 minGPT-6.1 Sol · Codex: final S = 0.620 at 447 min

Showing 2 of 2 runs. Hover a line or a model to isolate it; click a model to hide or show it.

Elapsed time includes HFSS solver waits. RF scores are diagnostic: official scoring and qualification are still pending.

Final designs

Geometry and S-parameters for each run.

Each run's final geometry beside its simulated response. Select a figure to zoom.

  • Dual-band Filter-v0 · GPT-6.1 Sol geometry

    Final geometry · dimensions in mm

    Fit

    Dual-band Filter-v0 · GPT-6.1 Sol S-parameters

    S21 transmission and S11 reflection in dB

    Fit
    GPT-6.1 SolCodexS = 0.62

  • Dual-band Filter-v0 · Opus 5.5 geometry

    Final geometry · dimensions in mm

    Fit

    Dual-band Filter-v0 · Opus 5.5 S-parameters

    S21 transmission and S11 reflection in dB

    Fit
    Opus 5.5Claude CodeS = 0.57