Add final-response verifier checkpoint

hf_model/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "[MASK]": 128000
+}

hf_model/artifact_manifest.json

@@ -0,0 +1,47 @@
+{
+  "artifact_schema_version": "final-response-verifier-artifact/v1",
+  "model_kind": "text-classification-cross-encoder",
+  "base_model": "microsoft/deberta-v3-small",
+  "label_mode": "production",
+  "input_schema_version": "final-response-verifier-input/v1",
+  "serializer": "serialize_final_response_state_v1",
+  "max_length": 768,
+  "requested_gpu_profile": "high_vram_quality",
+  "run_profile": "high_vram_quality",
+  "memory_profile": {
+    "max_length": 768,
+    "epochs": 5,
+    "train_batch_size": 16,
+    "eval_batch_size": 32,
+    "grad_accum": 4,
+    "max_per_label": 5000,
+    "force_retrain": false,
+    "export_cpu_only": true
+  },
+  "gpu_info": {
+    "available": true,
+    "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
+    "capability": [
+      12,
+      0
+    ],
+    "total_gb": 95.0
+  },
+  "precision_flags": {
+    "fp16": false,
+    "bf16": true,
+    "tf32": true
+  },
+  "onnx_file": "model.onnx",
+  "quantized_onnx_file": "model_quantized.onnx",
+  "labels": [
+    "valid_final_response",
+    "missing_tool_fact",
+    "contradicts_tool_result",
+    "unsupported_claim",
+    "failed_to_acknowledge_data_gap"
+  ],
+  "deployment_default": "shadow",
+  "shadow_first_reason": "experimental final-response verifier; promote only after eval replay",
+  "created_unix": 1779844717
+}

hf_model/config.json

@@ -0,0 +1,51 @@
+{
+  "architectures": [
+    "DebertaV2ForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 1,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "valid_final_response",
+    "1": "missing_tool_fact",
+    "2": "contradicts_tool_result",
+    "3": "unsupported_claim",
+    "4": "failed_to_acknowledge_data_gap"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "contradicts_tool_result": 2,
+    "failed_to_acknowledge_data_gap": 4,
+    "missing_tool_fact": 1,
+    "unsupported_claim": 3,
+    "valid_final_response": 0
+  },
+  "layer_norm_eps": 1e-07,
+  "legacy": true,
+  "max_position_embeddings": 512,
+  "max_relative_positions": -1,
+  "model_type": "deberta-v2",
+  "norm_rel_ebd": "layer_norm",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 6,
+  "pad_token_id": 0,
+  "pooler_dropout": 0,
+  "pooler_hidden_act": "gelu",
+  "pooler_hidden_size": 768,
+  "pos_att_type": [
+    "p2c",
+    "c2p"
+  ],
+  "position_biased_input": false,
+  "position_buckets": 256,
+  "relative_attention": true,
+  "share_att_key": true,
+  "transformers_version": "4.57.6",
+  "type_vocab_size": 0,
+  "vocab_size": 128100
+}

hf_model/input_schema.json

@@ -0,0 +1,134 @@
+{
+  "$id": "final-response-verifier-input/v1",
+  "type": "object",
+  "required": [
+    "schema_version",
+    "user_request",
+    "workflow_state",
+    "required_facts",
+    "tool_trace",
+    "tool_results",
+    "candidate_final_response"
+  ],
+  "properties": {
+    "schema_version": {
+      "const": "final-response-verifier-input/v1"
+    },
+    "user_request": {
+      "type": "string"
+    },
+    "workflow_state": {
+      "type": "object",
+      "required": [
+        "required_steps",
+        "completed_steps",
+        "pending_steps",
+        "terminal_tools"
+      ],
+      "properties": {
+        "required_steps": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          }
+        },
+        "completed_steps": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          }
+        },
+        "pending_steps": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          }
+        },
+        "terminal_tools": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          }
+        },
+        "recent_errors": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          }
+        }
+      },
+      "additionalProperties": false
+    },
+    "required_facts": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      }
+    },
+    "tool_trace": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      }
+    },
+    "tool_results": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "required": [
+          "tool_name",
+          "content"
+        ],
+        "properties": {
+          "tool_name": {
+            "type": "string"
+          },
+          "content": {
+            "type": "string"
+          }
+        },
+        "additionalProperties": true
+      }
+    },
+    "candidate_final_response": {
+      "type": "string"
+    },
+    "metadata": {
+      "type": "object",
+      "properties": {
+        "scenario_family": {
+          "type": [
+            "string",
+            "null"
+          ]
+        },
+        "requires_transform": {
+          "type": [
+            "boolean",
+            "null"
+          ]
+        },
+        "requires_synthesis": {
+          "type": [
+            "boolean",
+            "null"
+          ]
+        },
+        "requires_all_tool_facts": {
+          "type": [
+            "boolean",
+            "null"
+          ]
+        },
+        "must_acknowledge_missing_data": {
+          "type": [
+            "boolean",
+            "null"
+          ]
+        }
+      },
+      "additionalProperties": false
+    }
+  },
+  "additionalProperties": false
+}

hf_model/labels.json

@@ -0,0 +1,24 @@
+{
+  "label_mode": "production",
+  "labels": [
+    "valid_final_response",
+    "missing_tool_fact",
+    "contradicts_tool_result",
+    "unsupported_claim",
+    "failed_to_acknowledge_data_gap"
+  ],
+  "label2id": {
+    "valid_final_response": 0,
+    "missing_tool_fact": 1,
+    "contradicts_tool_result": 2,
+    "unsupported_claim": 3,
+    "failed_to_acknowledge_data_gap": 4
+  },
+  "id2label": {
+    "0": "valid_final_response",
+    "1": "missing_tool_fact",
+    "2": "contradicts_tool_result",
+    "3": "unsupported_claim",
+    "4": "failed_to_acknowledge_data_gap"
+  }
+}

hf_model/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e02866eaeff9f760e91414b51dc69371420ae83451c2ba06603d69affeb2b920
+size 567607780

hf_model/onnx_parity_report.json

@@ -0,0 +1,10 @@
+{
+  "schema_version": "final-response-verifier-onnx-parity/v1",
+  "rows": 10,
+  "pt_fp32_top_label_agreement": 1.0,
+  "pt_fp32_max_abs_diff": 2.980232238769531e-07,
+  "quantized_present": true,
+  "fp32_quantized_top_label_agreement": 1.0,
+  "fp32_quantized_disagreements": 0,
+  "fp32_quantized_max_abs_diff": 0.017383113503456116
+}

hf_model/special_tokens_map.json

@@ -0,0 +1,15 @@
+{
+  "bos_token": "[CLS]",
+  "cls_token": "[CLS]",
+  "eos_token": "[SEP]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": {
+    "content": "[UNK]",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}

hf_model/spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c679fbf93643d19aab7ee10c0b99e460bdbc02fedf34b92b05af343b4af586fd
+size 2464616

hf_model/thresholds.json

@@ -0,0 +1,32 @@
+{
+  "schema_version": "final-response-verifier-thresholds/v1",
+  "mode": "shadow",
+  "default_action": "allow",
+  "labels": {
+    "valid_final_response": {
+      "action": "allow",
+      "advisory_min_confidence": 0.0,
+      "enforce_min_confidence": 1.01
+    },
+    "missing_tool_fact": {
+      "action": "advisory_then_enforce_after_eval",
+      "advisory_min_confidence": 0.9,
+      "enforce_min_confidence": 0.995
+    },
+    "contradicts_tool_result": {
+      "action": "advisory_then_enforce_after_eval",
+      "advisory_min_confidence": 0.9,
+      "enforce_min_confidence": 0.995
+    },
+    "unsupported_claim": {
+      "action": "advisory_then_enforce_after_eval",
+      "advisory_min_confidence": 0.9,
+      "enforce_min_confidence": 0.995
+    },
+    "failed_to_acknowledge_data_gap": {
+      "action": "advisory_then_enforce_after_eval",
+      "advisory_min_confidence": 0.9,
+      "enforce_min_confidence": 0.995
+    }
+  }
+}

hf_model/tokenizer_config.json

@@ -0,0 +1,59 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "128000": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "[CLS]",
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "[CLS]",
+  "do_lower_case": false,
+  "eos_token": "[SEP]",
+  "extra_special_tokens": {},
+  "mask_token": "[MASK]",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "sp_model_kwargs": {},
+  "split_by_punct": false,
+  "tokenizer_class": "DebertaV2Tokenizer",
+  "unk_token": "[UNK]",
+  "vocab_type": "spm"
+}

hf_model/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf1f8fa2877d9aa33073fe14892601f04518942720c585894ee7d015fa4012e9
+size 5905

hf_model/training_provenance.json

@@ -0,0 +1,58 @@
+{
+  "schema_version": "final-response-verifier-training-provenance/v1",
+  "base_model": "microsoft/deberta-v3-small",
+  "run_profile": "high_vram_quality",
+  "gpu_info": {
+    "available": true,
+    "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
+    "capability": [
+      12,
+      0
+    ],
+    "total_gb": 95.0
+  },
+  "memory_profile": {
+    "max_length": 768,
+    "epochs": 5,
+    "train_batch_size": 16,
+    "eval_batch_size": 32,
+    "grad_accum": 4,
+    "max_per_label": 5000,
+    "force_retrain": false,
+    "export_cpu_only": true
+  },
+  "rows": 90,
+  "train_rows": 70,
+  "validation_rows": 10,
+  "test_rows": 10,
+  "label_counts": {
+    "contradicts_tool_result": 18,
+    "missing_tool_fact": 18,
+    "unsupported_claim": 18,
+    "valid_final_response": 18,
+    "failed_to_acknowledge_data_gap": 18
+  },
+  "resumed_from_checkpoint": true,
+  "train_metrics": {
+    "train_runtime": 14.31,
+    "train_samples_per_second": 24.458,
+    "train_steps_per_second": 0.699,
+    "total_flos": 16883336101500.0,
+    "train_loss": 1.6138598124186199,
+    "epoch": 3.0
+  },
+  "test_metrics": {
+    "eval_loss": 1.6188628673553467,
+    "eval_accuracy": 0.2,
+    "eval_macro_precision": 0.04,
+    "eval_macro_recall": 0.2,
+    "eval_macro_f1": 0.06666666666666667,
+    "eval_macro_precision_all_labels": 0.04,
+    "eval_macro_recall_all_labels": 0.2,
+    "eval_macro_f1_all_labels": 0.06666666666666667,
+    "eval_runtime": 0.3699,
+    "eval_samples_per_second": 27.032,
+    "eval_steps_per_second": 2.703,
+    "epoch": 3.0
+  }
+}