{
  "schema_version": 2,
  "manifest_ref": "reproduction:flvm",
  "status": "completed_public_diagnostic",
  "diagnostic_only": true,
  "paper_result_reproduced": false,
  "seeds": [
    42,
    43,
    44
  ],
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42,
      43,
      44
    ],
    "formal_comparison": false,
    "claim_policy": "public proxy mechanisms only; sparse negative signal and no production satisfaction labels"
  },
  "provenance": {
    "artifact_path": "docs/reproductions/2609.32839-flvm/metrics/public-seeds42-44.json",
    "dataset_fingerprint": "429e3b948828942e572f2c3a5be5a25799ffe75591d22d18cf417b9b534d31fd"
  },
  "runs": [
    {
      "paper": "2609.32839",
      "seed": 42,
      "steps_per_variant": 300,
      "dataset": "KuaiRand-Pure",
      "source_sha256": "429e3b948828942e572f2c3a5be5a25799ffe75591d22d18cf417b9b534d31fd",
      "split": {
        "train": 206847,
        "validation": 44325,
        "test": 44325,
        "method": "global chronological 70/15/15"
      },
      "variants": {
        "flvm": {
          "train_final_loss": 0.4424484074115753,
          "validation": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13857,
                "average_precision": 0.31605682325726536
              },
              "is_click": {
                "positive": 19648,
                "average_precision": 0.4600577007202335
              },
              "is_like": {
                "positive": 708,
                "average_precision": 0.01733348910047795
              },
              "is_hate": {
                "positive": 42,
                "average_precision": 0.001175510808045246
              }
            },
            "valence_mean_like_minus_hate": -0.15364575386047363
          },
          "test": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13707,
                "average_precision": 0.3111645636851855
              },
              "is_click": {
                "positive": 19472,
                "average_precision": 0.4479436991624803
              },
              "is_like": {
                "positive": 755,
                "average_precision": 0.01779009165517246
              },
              "is_hate": {
                "positive": 34,
                "average_precision": 0.010635348535999919
              }
            },
            "valence_mean_like_minus_hate": -0.03358817100524902
          }
        },
        "multitask_control": {
          "train_final_loss": 0.15926000475883484,
          "validation": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13857,
                "average_precision": 0.31728739997037103
              },
              "is_click": {
                "positive": 19648,
                "average_precision": 0.4524158359705734
              },
              "is_like": {
                "positive": 708,
                "average_precision": 0.01813240137555017
              },
              "is_hate": {
                "positive": 42,
                "average_precision": 0.0014829575034017305
              }
            }
          },
          "test": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13707,
                "average_precision": 0.3149439683677181
              },
              "is_click": {
                "positive": 19472,
                "average_precision": 0.44605617367378364
              },
              "is_like": {
                "positive": 755,
                "average_precision": 0.018446689521838776
              },
              "is_hate": {
                "positive": 34,
                "average_precision": 0.001651485714629451
              }
            }
          }
        }
      },
      "scope": "Public-data concept diagnostic, not YouTube Shorts satisfaction or online A/B."
    },
    {
      "paper": "2609.32839",
      "seed": 43,
      "steps_per_variant": 300,
      "dataset": "KuaiRand-Pure",
      "source_sha256": "429e3b948828942e572f2c3a5be5a25799ffe75591d22d18cf417b9b534d31fd",
      "split": {
        "train": 206847,
        "validation": 44325,
        "test": 44325,
        "method": "global chronological 70/15/15"
      },
      "variants": {
        "flvm": {
          "train_final_loss": 0.3981071412563324,
          "validation": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13857,
                "average_precision": 0.3194691178338808
              },
              "is_click": {
                "positive": 19648,
                "average_precision": 0.45872584199105443
              },
              "is_like": {
                "positive": 708,
                "average_precision": 0.021251420177759817
              },
              "is_hate": {
                "positive": 42,
                "average_precision": 0.0011259375381507516
              }
            },
            "valence_mean_like_minus_hate": 0.04744899272918701
          },
          "test": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13707,
                "average_precision": 0.3093646735106628
              },
              "is_click": {
                "positive": 19472,
                "average_precision": 0.4461575850572147
              },
              "is_like": {
                "positive": 755,
                "average_precision": 0.023549965710726383
              },
              "is_hate": {
                "positive": 34,
                "average_precision": 0.001223791939244535
              }
            },
            "valence_mean_like_minus_hate": 0.05599403381347656
          }
        },
        "multitask_control": {
          "train_final_loss": 0.1917411983013153,
          "validation": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13857,
                "average_precision": 0.33181863785882876
              },
              "is_click": {
                "positive": 19648,
                "average_precision": 0.46312592563403826
              },
              "is_like": {
                "positive": 708,
                "average_precision": 0.017664358273447932
              },
              "is_hate": {
                "positive": 42,
                "average_precision": 0.0010536861021558468
              }
            }
          },
          "test": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13707,
                "average_precision": 0.3251630760009834
              },
              "is_click": {
                "positive": 19472,
                "average_precision": 0.46376276903407004
              },
              "is_like": {
                "positive": 755,
                "average_precision": 0.018008183578231623
              },
              "is_hate": {
                "positive": 34,
                "average_precision": 0.0012677108686924856
              }
            }
          }
        }
      },
      "scope": "Public-data concept diagnostic, not YouTube Shorts satisfaction or online A/B."
    },
    {
      "paper": "2609.32839",
      "seed": 44,
      "steps_per_variant": 300,
      "dataset": "KuaiRand-Pure",
      "source_sha256": "429e3b948828942e572f2c3a5be5a25799ffe75591d22d18cf417b9b534d31fd",
      "split": {
        "train": 206847,
        "validation": 44325,
        "test": 44325,
        "method": "global chronological 70/15/15"
      },
      "variants": {
        "flvm": {
          "train_final_loss": 0.3813537061214447,
          "validation": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13857,
                "average_precision": 0.3166764700850034
              },
              "is_click": {
                "positive": 19648,
                "average_precision": 0.4522925493011909
              },
              "is_like": {
                "positive": 708,
                "average_precision": 0.019526893383527005
              },
              "is_hate": {
                "positive": 42,
                "average_precision": 0.0010143168648779298
              }
            },
            "valence_mean_like_minus_hate": 0.00898587703704834
          },
          "test": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13707,
                "average_precision": 0.300061621042743
              },
              "is_click": {
                "positive": 19472,
                "average_precision": 0.4349558683489409
              },
              "is_like": {
                "positive": 755,
                "average_precision": 0.020905141649434657
              },
              "is_hate": {
                "positive": 34,
                "average_precision": 0.008355568654159568
              }
            },
            "valence_mean_like_minus_hate": -0.0024628639221191406
          }
        },
        "multitask_control": {
          "train_final_loss": 0.17352968454360962,
          "validation": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13857,
                "average_precision": 0.3256517816999606
              },
              "is_click": {
                "positive": 19648,
                "average_precision": 0.4661139027068126
              },
              "is_like": {
                "positive": 708,
                "average_precision": 0.01809976850331222
              },
              "is_hate": {
                "positive": 42,
                "average_precision": 0.0011369043524293379
              }
            }
          },
          "test": {
            "impressions": 44325,
            "signals": {
              "long_view": {
                "positive": 13707,
                "average_precision": 0.3203128248452783
              },
              "is_click": {
                "positive": 19472,
                "average_precision": 0.45762509745865537
              },
              "is_like": {
                "positive": 755,
                "average_precision": 0.019055786438287592
              },
              "is_hate": {
                "positive": 34,
                "average_precision": 0.0014256358317382824
              }
            }
          }
        }
      },
      "scope": "Public-data concept diagnostic, not YouTube Shorts satisfaction or online A/B."
    }
  ]
}
