{
 "benchmark": "PDAGENT-BENCH",
 "categories": [
  {
   "key": "foundational",
   "name": "Foundational Knowledge",
   "short": "Foundational",
   "n": 90,
   "grader": "judge",
   "group": "Understanding",
   "desc": "Open-ended physical-design concepts: floorplanning, CTS, STA, DFT, extraction, device physics, flow I/O."
  },
  {
   "key": "root_cause",
   "name": "Root-Cause Analysis",
   "short": "Root-Cause",
   "n": 21,
   "grader": "judge",
   "group": "Understanding",
   "desc": "Diagnose real backend failures (hold fixing, IR drop, congestion, SDC, multi-voltage) and propose fixes."
  },
  {
   "key": "report",
   "name": "Report Comprehension",
   "short": "Report",
   "n": 11,
   "grader": "judge",
   "group": "Understanding",
   "desc": "Read EDA tool reports and screenshots (P&R and timing sign-off tools) and explain what they show. Multimodal."
  },
  {
   "key": "sta",
   "name": "Static Timing Analysis",
   "short": "STA",
   "n": 11,
   "grader": "rule",
   "group": "Understanding",
   "desc": "Timing calculations and multiple-choice STA questions, several from circuit diagrams. Multimodal."
  },
  {
   "key": "fm",
   "name": "Formal Verification",
   "short": "FM",
   "n": 22,
   "grader": "rule",
   "group": "Script Generation",
   "desc": "Exact Formality commands for library setup, read, match/verify and reporting."
  },
  {
   "key": "eco",
   "name": "ECO",
   "short": "ECO",
   "n": 10,
   "grader": "rule",
   "group": "Script Generation",
   "desc": "PrimeTime ECO commands: timing fixes, false paths, change lists, violation reports."
  },
  {
   "key": "pr_vendor1",
   "name": "P&R Vendor 1",
   "short": "P&R Vendor 1",
   "n": 88,
   "grader": "rule",
   "group": "Script Generation",
   "desc": "Exact Tcl for a commercial place-and-route tool: setup, floorplan, power, placement, CTS, routing, EM and antenna."
  },
  {
   "key": "pr_vendor2",
   "name": "P&R Vendor 2",
   "short": "P&R Vendor 2",
   "n": 90,
   "grader": "rule",
   "group": "Script Generation",
   "desc": "Exact Tcl for a second commercial place-and-route tool: initialization and MMMC through floorplan, CTS, ECO and sign-off export."
  }
 ],
 "rounds": {
  "r1": {
   "label": "Earlier evaluation",
   "date": null,
   "judge": "LLM judge + human evaluation (paper Table 1)",
   "settings": "temperature 0 (pass@1) / 0.8 (pass@5), provider default output caps"
  },
  "r2": {
   "label": "2026-09-23 evaluation",
   "date": "2026-09-23",
   "judge": "GPT-5.4-mini rubric judge (blind)",
   "settings": "reasoning effort medium, no temperature, 32K output cap"
  }
 },
 "models": [
  {
   "name": "Qwen-VL-8B",
   "provider": "Alibaba",
   "open_weights": true,
   "multimodal": true,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.533,
     "p5": 0.573,
     "p1_human": 0.52,
     "p5_human": 0.558
    },
    "root_cause": {
     "p1": 0.245,
     "p5": 0.319,
     "p1_human": 0.228,
     "p5_human": 0.269
    },
    "report": {
     "p1": 0.327,
     "p5": 0.464,
     "p1_human": 0.293,
     "p5_human": 0.421
    },
    "sta": {
     "p1": 0.545,
     "p5": 0.636
    },
    "fm": {
     "p1": 0.273,
     "p5": 0.455
    },
    "eco": {
     "p1": 0.1,
     "p5": 0.1
    },
    "pr_vendor1": {
     "p1": 0.068,
     "p5": 0.102
    },
    "pr_vendor2": {
     "p1": 0.122,
     "p5": 0.133
    }
   },
   "overall": {
    "p1": 0.2766,
    "p5": 0.3478
   },
   "n_categories": 8
  },
  {
   "name": "Qwen-VL-30B",
   "provider": "Alibaba",
   "open_weights": true,
   "multimodal": true,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.692,
     "p5": 0.695,
     "p1_human": 0.661,
     "p5_human": 0.679
    },
    "root_cause": {
     "p1": 0.379,
     "p5": 0.46,
     "p1_human": 0.368,
     "p5_human": 0.4
    },
    "report": {
     "p1": 0.323,
     "p5": 0.427,
     "p1_human": 0.294,
     "p5_human": 0.415
    },
    "sta": {
     "p1": 0.545,
     "p5": 0.636
    },
    "fm": {
     "p1": 0.273,
     "p5": 0.545
    },
    "eco": {
     "p1": 0.1,
     "p5": 0.1
    },
    "pr_vendor1": {
     "p1": 0.091,
     "p5": 0.125
    },
    "pr_vendor2": {
     "p1": 0.111,
     "p5": 0.178
    }
   },
   "overall": {
    "p1": 0.3143,
    "p5": 0.3957
   },
   "n_categories": 8
  },
  {
   "name": "Qwen3-8B",
   "provider": "Alibaba",
   "open_weights": true,
   "multimodal": false,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.543,
     "p5": 0.585,
     "p1_human": 0.488,
     "p5_human": 0.573
    },
    "root_cause": {
     "p1": 0.255,
     "p5": 0.286,
     "p1_human": 0.216,
     "p5_human": 0.263
    },
    "fm": {
     "p1": 0.273,
     "p5": 0.5
    },
    "eco": {
     "p1": 0.0,
     "p5": 0.0
    },
    "pr_vendor1": {
     "p1": 0.08,
     "p5": 0.102
    },
    "pr_vendor2": {
     "p1": 0.122,
     "p5": 0.144
    }
   },
   "overall": {
    "p1": 0.2122,
    "p5": 0.2695
   },
   "n_categories": 6
  },
  {
   "name": "DeepSeek-Chat",
   "provider": "DeepSeek",
   "open_weights": true,
   "multimodal": false,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.726,
     "p5": 0.799,
     "p1_human": 0.689,
     "p5_human": 0.773
    },
    "root_cause": {
     "p1": 0.486,
     "p5": 0.521,
     "p1_human": 0.434,
     "p5_human": 0.473
    },
    "fm": {
     "p1": 0.409,
     "p5": 0.591
    },
    "eco": {
     "p1": 0.3,
     "p5": 0.5
    },
    "pr_vendor1": {
     "p1": 0.34,
     "p5": 0.364
    },
    "pr_vendor2": {
     "p1": 0.256,
     "p5": 0.356
    }
   },
   "overall": {
    "p1": 0.4195,
    "p5": 0.5218
   },
   "n_categories": 6
  },
  {
   "name": "DeepSeek-Reasoner",
   "provider": "DeepSeek",
   "open_weights": true,
   "multimodal": false,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.737,
     "p5": 0.806,
     "p1_human": 0.684,
     "p5_human": 0.788
    },
    "root_cause": {
     "p1": 0.612,
     "p5": 0.68,
     "p1_human": 0.585,
     "p5_human": 0.63
    },
    "fm": {
     "p1": 0.455,
     "p5": 0.591
    },
    "eco": {
     "p1": 0.3,
     "p5": 0.5
    },
    "pr_vendor1": {
     "p1": 0.216,
     "p5": 0.352
    },
    "pr_vendor2": {
     "p1": 0.233,
     "p5": 0.356
    }
   },
   "overall": {
    "p1": 0.4255,
    "p5": 0.5475
   },
   "n_categories": 6
  },
  {
   "name": "GPT-5-mini",
   "provider": "OpenAI",
   "open_weights": false,
   "multimodal": true,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.8,
     "p5": 0.82,
     "p1_human": 0.701,
     "p5_human": 0.79
    },
    "root_cause": {
     "p1": 0.56,
     "p5": 0.655,
     "p1_human": 0.54,
     "p5_human": 0.615
    },
    "report": {
     "p1": 0.655,
     "p5": 0.685,
     "p1_human": 0.607,
     "p5_human": 0.666
    },
    "sta": {
     "p1": 0.545,
     "p5": 0.636
    },
    "fm": {
     "p1": 0.227,
     "p5": 0.364
    },
    "eco": {
     "p1": 0.3,
     "p5": 0.3
    },
    "pr_vendor1": {
     "p1": 0.057,
     "p5": 0.114
    },
    "pr_vendor2": {
     "p1": 0.133,
     "p5": 0.189
    }
   },
   "overall": {
    "p1": 0.4096,
    "p5": 0.4704
   },
   "n_categories": 8
  },
  {
   "name": "GPT-5.1",
   "provider": "OpenAI",
   "open_weights": false,
   "multimodal": true,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.82,
     "p5": 0.84,
     "p1_human": 0.802,
     "p5_human": 0.83
    },
    "root_cause": {
     "p1": 0.655,
     "p5": 0.674,
     "p1_human": 0.619,
     "p5_human": 0.624
    },
    "report": {
     "p1": 0.655,
     "p5": 0.705,
     "p1_human": 0.604,
     "p5_human": 0.649
    },
    "sta": {
     "p1": 0.545,
     "p5": 0.545
    },
    "fm": {
     "p1": 0.545,
     "p5": 0.636
    },
    "eco": {
     "p1": 0.2,
     "p5": 0.2
    },
    "pr_vendor1": {
     "p1": 0.318,
     "p5": 0.432
    },
    "pr_vendor2": {
     "p1": 0.278,
     "p5": 0.311
    }
   },
   "overall": {
    "p1": 0.502,
    "p5": 0.5429
   },
   "n_categories": 8
  },
  {
   "name": "Claude Sonnet 4.6",
   "provider": "Anthropic",
   "open_weights": false,
   "multimodal": true,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.826,
     "p5": 0.835,
     "p1_human": 0.794,
     "p5_human": 0.821
    },
    "root_cause": {
     "p1": 0.688,
     "p5": 0.7,
     "p1_human": 0.66,
     "p5_human": 0.664
    },
    "report": {
     "p1": 0.668,
     "p5": 0.736,
     "p1_human": 0.625,
     "p5_human": 0.717
    },
    "sta": {
     "p1": 0.636,
     "p5": 0.636
    },
    "fm": {
     "p1": 0.636,
     "p5": 0.682
    },
    "eco": {
     "p1": 0.5,
     "p5": 0.5
    },
    "pr_vendor1": {
     "p1": 0.341,
     "p5": 0.352
    },
    "pr_vendor2": {
     "p1": 0.289,
     "p5": 0.311
    }
   },
   "overall": {
    "p1": 0.573,
    "p5": 0.594
   },
   "n_categories": 8
  },
  {
   "name": "Claude Opus 4.7",
   "provider": "Anthropic",
   "open_weights": false,
   "multimodal": true,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.824,
     "p5": 0.842,
     "p1_human": 0.784,
     "p5_human": 0.829
    },
    "root_cause": {
     "p1": 0.762,
     "p5": 0.767,
     "p1_human": 0.714,
     "p5_human": 0.731
    },
    "report": {
     "p1": 0.641,
     "p5": 0.659,
     "p1_human": 0.614,
     "p5_human": 0.629
    },
    "sta": {
     "p1": 0.727,
     "p5": 0.727
    },
    "fm": {
     "p1": 0.727,
     "p5": 0.727
    },
    "eco": {
     "p1": 0.4,
     "p5": 0.5
    },
    "pr_vendor1": {
     "p1": 0.341,
     "p5": 0.375
    },
    "pr_vendor2": {
     "p1": 0.367,
     "p5": 0.411
    }
   },
   "overall": {
    "p1": 0.5986,
    "p5": 0.626
   },
   "n_categories": 8
  },
  {
   "name": "GPT-5.5",
   "provider": "OpenAI",
   "open_weights": false,
   "multimodal": true,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.804,
     "p5": 0.83,
     "p1_human": 0.778,
     "p5_human": 0.807
    },
    "root_cause": {
     "p1": 0.733,
     "p5": 0.774,
     "p1_human": 0.689,
     "p5_human": 0.732
    },
    "report": {
     "p1": 0.746,
     "p5": 0.75,
     "p1_human": 0.72,
     "p5_human": 0.732
    },
    "sta": {
     "p1": 0.727,
     "p5": 0.727
    },
    "fm": {
     "p1": 0.818,
     "p5": 0.864
    },
    "eco": {
     "p1": 0.5,
     "p5": 0.6
    },
    "pr_vendor1": {
     "p1": 0.455,
     "p5": 0.58
    },
    "pr_vendor2": {
     "p1": 0.422,
     "p5": 0.511
    }
   },
   "overall": {
    "p1": 0.6506,
    "p5": 0.7045
   },
   "n_categories": 8
  },
  {
   "name": "GPT-5.4-mini",
   "provider": "OpenAI",
   "open_weights": false,
   "multimodal": true,
   "round": "r1",
   "scores": {
    "foundational": {
     "p1": 0.723,
     "p5": 0.741,
     "p1_human": 0.668,
     "p5_human": 0.726
    },
    "root_cause": {
     "p1": 0.545,
     "p5": 0.581,
     "p1_human": 0.502,
     "p5_human": 0.526
    },
    "report": {
     "p1": 0.461,
     "p5": 0.514,
     "p1_human": 0.422,
     "p5_human": 0.487
    },
    "sta": {
     "p1": 0.636,
     "p5": 0.636
    },
    "fm": {
     "p1": 0.591,
     "p5": 0.727
    },
    "eco": {
     "p1": 0.5,
     "p5": 0.5
    },
    "pr_vendor1": {
     "p1": 0.295,
     "p5": 0.398
    },
    "pr_vendor2": {
     "p1": 0.267,
     "p5": 0.356
    }
   },
   "overall": {
    "p1": 0.5023,
    "p5": 0.5566
   },
   "n_categories": 8
  },
  {
   "name": "DeepSeek-V4.1-Flash",
   "provider": "DeepSeek",
   "open_weights": true,
   "multimodal": true,
   "api_model": "deepseek-flash",
   "round": "r2",
   "scores": {
    "foundational": {
     "p1": 0.8483,
     "p5": 0.8634,
     "p5_best": 0.93
    },
    "root_cause": {
     "p1": 0.669,
     "p5": 0.659,
     "p5_best": 0.7738
    },
    "report": {
     "p1": 0.5045,
     "p5": 0.5664,
     "p5_best": 0.7045
    },
    "sta": {
     "p1": 0.7273,
     "p5": 0.7273
    },
    "fm": {
     "p1": 0.7727,
     "p5": 0.9545
    },
    "eco": {
     "p1": 0.4,
     "p5": 0.5
    },
    "pr_vendor1": {
     "p1": 0.3636,
     "p5": 0.4659
    },
    "pr_vendor2": {
     "p1": 0.3222,
     "p5": 0.4444
    }
   },
   "overall": {
    "p1": 0.5759,
    "p5": 0.6476
   },
   "n_categories": 8
  },
  {
   "name": "GPT-5.6 Luna",
   "provider": "OpenAI",
   "open_weights": false,
   "multimodal": true,
   "api_model": "gpt-5.6-luna",
   "round": "r2",
   "scores": {
    "foundational": {
     "p1": 0.8483,
     "p5": 0.8294,
     "p5_best": 0.8939
    },
    "root_cause": {
     "p1": 0.5762,
     "p5": 0.59,
     "p5_best": 0.7071
    },
    "report": {
     "p1": 0.6,
     "p5": 0.6036,
     "p5_best": 0.7182
    },
    "sta": {
     "p1": 0.4545,
     "p5": 0.5455
    },
    "fm": {
     "p1": 0.5909,
     "p5": 0.7727
    },
    "eco": {
     "p1": 0.6,
     "p5": 0.5
    },
    "pr_vendor1": {
     "p1": 0.3864,
     "p5": 0.4545
    },
    "pr_vendor2": {
     "p1": 0.3667,
     "p5": 0.4667
    }
   },
   "overall": {
    "p1": 0.5529,
    "p5": 0.5953
   },
   "n_categories": 8
  },
  {
   "name": "GPT-5.6 Sol",
   "provider": "OpenAI",
   "open_weights": false,
   "multimodal": true,
   "api_model": "gpt-5.6-sol",
   "round": "r2",
   "scores": {
    "foundational": {
     "p1": 0.8567,
     "p5": 0.8706,
     "p5_best": 0.9294
    },
    "root_cause": {
     "p1": 0.6571,
     "p5": 0.6714,
     "p5_best": 0.7786
    },
    "report": {
     "p1": 0.6818,
     "p5": 0.6582,
     "p5_best": 0.7773
    },
    "sta": {
     "p1": 0.7273,
     "p5": 0.6364
    },
    "fm": {
     "p1": 0.8182,
     "p5": 0.8182
    },
    "eco": {
     "p1": 0.4,
     "p5": 0.4
    },
    "pr_vendor1": {
     "p1": 0.5227,
     "p5": 0.6023
    },
    "pr_vendor2": {
     "p1": 0.4111,
     "p5": 0.5111
    }
   },
   "overall": {
    "p1": 0.6344,
    "p5": 0.646
   },
   "n_categories": 8
  },
  {
   "name": "Claude Sonnet 5",
   "provider": "Anthropic",
   "open_weights": false,
   "multimodal": true,
   "api_model": "claude-sonnet-5",
   "round": "r2",
   "scores": {
    "foundational": {
     "p1": 0.8289,
     "p5": 0.8319,
     "p5_best": 0.9028
    },
    "root_cause": {
     "p1": 0.5738,
     "p5": 0.5786,
     "p5_best": 0.7238
    },
    "report": {
     "p1": 0.6227,
     "p5": 0.6082,
     "p5_best": 0.7182
    },
    "sta": {
     "p1": 0.5455,
     "p5": 0.7273
    },
    "fm": {
     "p1": 0.5455,
     "p5": 0.6364
    },
    "eco": {
     "p1": 0.4,
     "p5": 0.6
    },
    "pr_vendor1": {
     "p1": 0.2727,
     "p5": 0.3409
    },
    "pr_vendor2": {
     "p1": 0.3333,
     "p5": 0.3667
    }
   },
   "overall": {
    "p1": 0.5153,
    "p5": 0.5862
   },
   "n_categories": 8
  },
  {
   "name": "Claude Opus 5",
   "provider": "Anthropic",
   "open_weights": false,
   "multimodal": true,
   "api_model": "claude-opus-5",
   "round": "r2",
   "scores": {
    "foundational": {
     "p1": 0.9,
     "p5": 0.9079,
     "p5_best": 0.96
    },
    "root_cause": {
     "p1": 0.7476,
     "p5": 0.7852,
     "p5_best": 0.8786
    },
    "report": {
     "p1": 0.5364,
     "p5": 0.6836,
     "p5_best": 0.8091
    },
    "sta": {
     "p1": 0.6364,
     "p5": 0.7273
    },
    "fm": {
     "p1": 0.8182,
     "p5": 0.9091
    },
    "eco": {
     "p1": 0.4,
     "p5": 0.4
    },
    "pr_vendor1": {
     "p1": 0.3977,
     "p5": 0.5341
    },
    "pr_vendor2": {
     "p1": 0.4,
     "p5": 0.4444
    }
   },
   "overall": {
    "p1": 0.6045,
    "p5": 0.674
   },
   "n_categories": 8
  }
 ]
}
