Instructions to use veriloop-lab/veriloop-coder-e1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use veriloop-lab/veriloop-coder-e1 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="veriloop-lab/veriloop-coder-e1")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("veriloop-lab/veriloop-coder-e1")
model = AutoModelForImageTextToText.from_pretrained("veriloop-lab/veriloop-coder-e1")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

PEFT
How to use veriloop-lab/veriloop-coder-e1 with PEFT:
```
Task type is invalid.
```
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use veriloop-lab/veriloop-coder-e1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "veriloop-lab/veriloop-coder-e1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "veriloop-lab/veriloop-coder-e1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/veriloop-lab/veriloop-coder-e1

SGLang

How to use veriloop-lab/veriloop-coder-e1 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "veriloop-lab/veriloop-coder-e1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "veriloop-lab/veriloop-coder-e1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "veriloop-lab/veriloop-coder-e1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "veriloop-lab/veriloop-coder-e1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use veriloop-lab/veriloop-coder-e1 with Docker Model Runner:
```
docker model run hf.co/veriloop-lab/veriloop-coder-e1
```

ConorWang commited on Apr 17

Commit

a856587

verified ·

1 Parent(s): 8ad54e3

Upload uncertainty adapter artifacts to repo root

Browse files

Files changed (12) hide show

adapter/adapter_config.json +2 -11
adapter/adapter_model.safetensors +2 -2
best_checkpoint_manifest.json +27 -21
epoch_history.json +104 -108
host_manifest.json +24 -114
tokenizer/tokenizer.json +1 -1
uncertainty_adapter_plan.json +469 -0
uncertainty_adapter_train_result.json +564 -0
uncertainty_eval.jsonl +55 -0
uncertainty_head.pt +3 -0
uncertainty_train.jsonl +0 -0
uncertainty_training_manifest.json +105 -0

adapter/adapter_config.json CHANGED Viewed

@@ -30,17 +30,8 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "surface_host.toolspec_bridge.adapter",
-    "surface_host.validator_feedback_bridge.adapter",
-    "surface_host.toolspec_head.receipt_formatter",
-    "surface_host.toolspec_head.precondition_adapter",
-    "surface_host.toolspec_head.validator_gate",
-    "surface_host.toolspec_head.postcondition_adapter",
-    "surface_host.tool_protocol_adapter.bridge",
-    "surface_host.runtime_harness_adapter.bridge",
-    "surface_host.toolspec_head.trigger_gate",
-    "surface_host.request_normalizer.adapter",
-    "surface_host.toolspec_head.param_schema_adapter"
   ],
   "target_parameters": null,
   "task_type": "FEATURE_EXTRACTION",

   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "surface_host.uncertainty_head.proj",
+    "surface_host.uncertainty_head.calibration_mlp"
   ],
   "target_parameters": null,
   "task_type": "FEATURE_EXTRACTION",

adapter/adapter_model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e62775d0cd91a0604b237b7d50cdb1da249ec85e8f9a2c6ad4fd7ed8afc805a2
-size 724048

 version https://git-lfs.github.com/spec/v1
+oid sha256:fcca67011c0b0b0eaa44fc4826f4908bdd6ca8d82b192916a6e35556e292fa7f
+size 98968

best_checkpoint_manifest.json CHANGED Viewed

@@ -1,27 +1,33 @@
 {
-  "best_epoch": 4,
-  "best_quality_score": 0.46349615617231893,
   "eval_metrics": {
-    "avg_binary_accuracy": 0.81,
-    "confirmation_required_accuracy": 0.8,
-    "count": 25,
-    "eval_batches": 25,
-    "eval_loss": 6.897225952148437,
-    "mode_accuracy": 0.56,
-    "permission_bit_accuracy": 0.8545454545454545,
-    "precondition_ok_accuracy": 0.76,
-    "quality_score": 0.46349615617231893,
-    "rollback_supported_accuracy": 0.8,
-    "schema_ok_accuracy": 0.76,
-    "session_required_accuracy": 1.0,
-    "tool_accuracy": 0.44,
-    "trigger_accuracy": 0.76,
-    "validator_required_accuracy": 0.8,
-    "worktree_required_accuracy": 0.8
   },
   "train_metrics": {
-    "loss": 0.42175399448190415,
-    "micro_batches": 70,
-    "optimizer_steps": 5
   }
 }

 {
+  "best_epoch": 2,
+  "best_quality_score": 0.8625935807221907,
   "eval_metrics": {
+    "count": 55,
+    "mae": {
+      "u_answer": 0.15174226462841034,
+      "u_evidence": 0.19610758125782013,
+      "u_exec": 0.18561214208602905,
+      "u_risk": 0.1553734689950943,
+      "u_spec": 0.21633382141590118
+    },
+    "mean_mae": 0.18103384971618652,
+    "mean_rmse": 0.24169571697711945,
+    "moderate_accuracy": 0.6727272727272727,
+    "quality_score": 0.8625935807221907,
+    "rmse": {
+      "u_answer": 0.18724055588245392,
+      "u_evidence": 0.22527915239334106,
+      "u_exec": 0.25238174200057983,
+      "u_risk": 0.20667441189289093,
+      "u_spec": 0.3369026482105255
+    },
+    "tight_accuracy": 0.4,
+    "weighted_mae": 0.18083095811830807,
+    "weighted_rmse": 0.24125460771003793
   },
   "train_metrics": {
+    "loss": 0.009006613283418119,
+    "micro_batches": 154,
+    "optimizer_steps": 10
   }
 }

epoch_history.json CHANGED Viewed

@@ -3,141 +3,137 @@
     {
       "epoch": 1,
       "eval_metrics": {
-        "avg_binary_accuracy": 0.795,
-        "confirmation_required_accuracy": 0.8,
-        "count": 25,
-        "eval_batches": 25,
-        "eval_loss": 7.704372234344483,
-        "mode_accuracy": 0.44,
-        "permission_bit_accuracy": 0.850909090909091,
-        "precondition_ok_accuracy": 0.76,
-        "quality_score": 0.21877532557574186,
-        "rollback_supported_accuracy": 0.72,
-        "schema_ok_accuracy": 0.76,
-        "session_required_accuracy": 1.0,
-        "tool_accuracy": 0.24,
-        "trigger_accuracy": 0.76,
-        "validator_required_accuracy": 0.8,
-        "worktree_required_accuracy": 0.76
       },
       "improved": true,
-      "quality_score": 0.21877532557574186,
       "train_metrics": {
-        "loss": 0.5388983534915107,
-        "micro_batches": 70,
-        "optimizer_steps": 5
       }
     },
     {
       "epoch": 2,
       "eval_metrics": {
-        "avg_binary_accuracy": 0.81,
-        "confirmation_required_accuracy": 0.8,
-        "count": 25,
-        "eval_batches": 25,
-        "eval_loss": 7.129253711700439,
-        "mode_accuracy": 0.48,
-        "permission_bit_accuracy": 0.8545454545454545,
-        "precondition_ok_accuracy": 0.76,
-        "quality_score": 0.4400157397790389,
-        "rollback_supported_accuracy": 0.8,
-        "schema_ok_accuracy": 0.76,
-        "session_required_accuracy": 1.0,
-        "tool_accuracy": 0.44,
-        "trigger_accuracy": 0.76,
-        "validator_required_accuracy": 0.8,
-        "worktree_required_accuracy": 0.8
       },
       "improved": true,
-      "quality_score": 0.4400157397790389,
       "train_metrics": {
-        "loss": 0.4655745736190251,
-        "micro_batches": 70,
-        "optimizer_steps": 5
       }
     },
     {
       "epoch": 3,
       "eval_metrics": {
-        "avg_binary_accuracy": 0.81,
-        "confirmation_required_accuracy": 0.8,
-        "count": 25,
-        "eval_batches": 25,
-        "eval_loss": 6.935494079589843,
-        "mode_accuracy": 0.48,
-        "permission_bit_accuracy": 0.8545454545454545,
-        "precondition_ok_accuracy": 0.76,
-        "quality_score": 0.4229221342606978,
-        "rollback_supported_accuracy": 0.8,
-        "schema_ok_accuracy": 0.76,
-        "session_required_accuracy": 1.0,
-        "tool_accuracy": 0.4,
-        "trigger_accuracy": 0.76,
-        "validator_required_accuracy": 0.8,
-        "worktree_required_accuracy": 0.8
       },
       "improved": false,
-      "quality_score": 0.4229221342606978,
       "train_metrics": {
-        "loss": 0.43205853487764084,
-        "micro_batches": 70,
-        "optimizer_steps": 5
       }
     },
     {
       "epoch": 4,
       "eval_metrics": {
-        "avg_binary_accuracy": 0.81,
-        "confirmation_required_accuracy": 0.8,
-        "count": 25,
-        "eval_batches": 25,
-        "eval_loss": 6.897225952148437,
-        "mode_accuracy": 0.56,
-        "permission_bit_accuracy": 0.8545454545454545,
-        "precondition_ok_accuracy": 0.76,
-        "quality_score": 0.46349615617231893,
-        "rollback_supported_accuracy": 0.8,
-        "schema_ok_accuracy": 0.76,
-        "session_required_accuracy": 1.0,
-        "tool_accuracy": 0.44,
-        "trigger_accuracy": 0.76,
-        "validator_required_accuracy": 0.8,
-        "worktree_required_accuracy": 0.8
-      },
-      "improved": true,
-      "quality_score": 0.46349615617231893,
-      "train_metrics": {
-        "loss": 0.42175399448190415,
-        "micro_batches": 70,
-        "optimizer_steps": 5
-      }
-    },
-    {
-      "epoch": 5,
-      "eval_metrics": {
-        "avg_binary_accuracy": 0.81,
-        "confirmation_required_accuracy": 0.8,
-        "count": 25,
-        "eval_batches": 25,
-        "eval_loss": 6.89590030670166,
-        "mode_accuracy": 0.56,
-        "permission_bit_accuracy": 0.8545454545454545,
-        "precondition_ok_accuracy": 0.76,
-        "quality_score": 0.4635160408540206,
-        "rollback_supported_accuracy": 0.8,
-        "schema_ok_accuracy": 0.76,
-        "session_required_accuracy": 1.0,
-        "tool_accuracy": 0.44,
-        "trigger_accuracy": 0.76,
-        "validator_required_accuracy": 0.8,
-        "worktree_required_accuracy": 0.8
       },
       "improved": false,
-      "quality_score": 0.4635160408540206,
       "train_metrics": {
-        "loss": 0.4198248211826597,
-        "micro_batches": 70,
-        "optimizer_steps": 5
       }
     }
   ]

     {
       "epoch": 1,
       "eval_metrics": {
+        "count": 55,
+        "mae": {
+          "u_answer": 0.1848176270723343,
+          "u_evidence": 0.20735961198806763,
+          "u_exec": 0.21166233718395233,
+          "u_risk": 0.1652287244796753,
+          "u_spec": 0.24522197246551514
+        },
+        "mean_mae": 0.2028580605983734,
+        "mean_rmse": 0.2640994191169739,
+        "moderate_accuracy": 0.6036363636363636,
+        "quality_score": 0.8344919736975306,
+        "rmse": {
+          "u_answer": 0.2176007330417633,
+          "u_evidence": 0.23360508680343628,
+          "u_exec": 0.2795017957687378,
+          "u_risk": 0.22334341704845428,
+          "u_spec": 0.36644604802131653
+        },
+        "tight_accuracy": 0.36727272727272725,
+        "weighted_mae": 0.20175925098418107,
+        "weighted_rmse": 0.26307498343893954
       },
       "improved": true,
+      "quality_score": 0.8344919736975306,
       "train_metrics": {
+        "loss": 0.011121059906239584,
+        "micro_batches": 154,
+        "optimizer_steps": 10
       }
     },
     {
       "epoch": 2,
       "eval_metrics": {
+        "count": 55,
+        "mae": {
+          "u_answer": 0.15174226462841034,
+          "u_evidence": 0.19610758125782013,
+          "u_exec": 0.18561214208602905,
+          "u_risk": 0.1553734689950943,
+          "u_spec": 0.21633382141590118
+        },
+        "mean_mae": 0.18103384971618652,
+        "mean_rmse": 0.24169571697711945,
+        "moderate_accuracy": 0.6727272727272727,
+        "quality_score": 0.8625935807221907,
+        "rmse": {
+          "u_answer": 0.18724055588245392,
+          "u_evidence": 0.22527915239334106,
+          "u_exec": 0.25238174200057983,
+          "u_risk": 0.20667441189289093,
+          "u_spec": 0.3369026482105255
+        },
+        "tight_accuracy": 0.4,
+        "weighted_mae": 0.18083095811830807,
+        "weighted_rmse": 0.24125460771003793
       },
       "improved": true,
+      "quality_score": 0.8625935807221907,
       "train_metrics": {
+        "loss": 0.009006613283418119,
+        "micro_batches": 154,
+        "optimizer_steps": 10
       }
     },
     {
       "epoch": 3,
       "eval_metrics": {
+        "count": 55,
+        "mae": {
+          "u_answer": 0.14970968663692474,
+          "u_evidence": 0.19498984515666962,
+          "u_exec": 0.18136857450008392,
+          "u_risk": 0.1565144807100296,
+          "u_spec": 0.21266111731529236
+        },
+        "mean_mae": 0.17904874682426453,
+        "mean_rmse": 0.23914317786693573,
+        "moderate_accuracy": 0.6545454545454545,
+        "quality_score": 0.8630735522772117,
+        "rmse": {
+          "u_answer": 0.1858142465353012,
+          "u_evidence": 0.22504529356956482,
+          "u_exec": 0.24751117825508118,
+          "u_risk": 0.20870783925056458,
+          "u_spec": 0.3286373019218445
+        },
+        "tight_accuracy": 0.3927272727272727,
+        "weighted_mae": 0.1789679336398722,
+        "weighted_rmse": 0.23888496180570432
       },
       "improved": false,
+      "quality_score": 0.8630735522772117,
       "train_metrics": {
+        "loss": 0.008164266918108544,
+        "micro_batches": 154,
+        "optimizer_steps": 10
       }
     },
     {
       "epoch": 4,
       "eval_metrics": {
+        "count": 55,
+        "mae": {
+          "u_answer": 0.1496635228395462,
+          "u_evidence": 0.19500812888145447,
+          "u_exec": 0.18158338963985443,
+          "u_risk": 0.1564789116382599,
+          "u_spec": 0.21253055334091187
+        },
+        "mean_mae": 0.1790529191493988,
+        "mean_rmse": 0.23915143311023712,
+        "moderate_accuracy": 0.6545454545454545,
+        "quality_score": 0.8630639342548476,
+        "rmse": {
+          "u_answer": 0.1857716143131256,
+          "u_evidence": 0.22511473298072815,
+          "u_exec": 0.24768365919589996,
+          "u_risk": 0.20870527625083923,
+          "u_spec": 0.32848188281059265
+        },
+        "tight_accuracy": 0.3927272727272727,
+        "weighted_mae": 0.17897483332915498,
+        "weighted_rmse": 0.2388972981611497
       },
       "improved": false,
+      "quality_score": 0.8630639342548476,
       "train_metrics": {
+        "loss": 0.008081083455086723,
+        "micro_batches": 154,
+        "optimizer_steps": 10
       }
     }
   ]

host_manifest.json CHANGED Viewed

@@ -68,16 +68,6 @@
     "query_runtime_engine.adapter",
     "request_normalizer",
     "request_normalizer.adapter",
-    "request_normalizer.adapter.base_layer",
-    "request_normalizer.adapter.lora_A",
-    "request_normalizer.adapter.lora_A.default",
-    "request_normalizer.adapter.lora_B",
-    "request_normalizer.adapter.lora_B.default",
-    "request_normalizer.adapter.lora_dropout",
-    "request_normalizer.adapter.lora_dropout.default",
-    "request_normalizer.adapter.lora_embedding_A",
-    "request_normalizer.adapter.lora_embedding_B",
-    "request_normalizer.adapter.lora_magnitude_vector",
     "rollback_adapter",
     "rollback_adapter.head",
     "rollback_engine",
@@ -86,16 +76,6 @@
     "runtime_binding.adapter",
     "runtime_harness_adapter",
     "runtime_harness_adapter.bridge",
-    "runtime_harness_adapter.bridge.base_layer",
-    "runtime_harness_adapter.bridge.lora_A",
-    "runtime_harness_adapter.bridge.lora_A.default",
-    "runtime_harness_adapter.bridge.lora_B",
-    "runtime_harness_adapter.bridge.lora_B.default",
-    "runtime_harness_adapter.bridge.lora_dropout",
-    "runtime_harness_adapter.bridge.lora_dropout.default",
-    "runtime_harness_adapter.bridge.lora_embedding_A",
-    "runtime_harness_adapter.bridge.lora_embedding_B",
-    "runtime_harness_adapter.bridge.lora_magnitude_vector",
     "runtime_harness_uncertainty_bridge",
     "runtime_harness_uncertainty_bridge.adapter",
     "sandbox_rollback_bridge",
@@ -107,110 +87,40 @@
     "session_state_manager.rollback_state",
     "tool_protocol_adapter",
     "tool_protocol_adapter.bridge",
-    "tool_protocol_adapter.bridge.base_layer",
-    "tool_protocol_adapter.bridge.lora_A",
-    "tool_protocol_adapter.bridge.lora_A.default",
-    "tool_protocol_adapter.bridge.lora_B",
-    "tool_protocol_adapter.bridge.lora_B.default",
-    "tool_protocol_adapter.bridge.lora_dropout",
-    "tool_protocol_adapter.bridge.lora_dropout.default",
-    "tool_protocol_adapter.bridge.lora_embedding_A",
-    "tool_protocol_adapter.bridge.lora_embedding_B",
-    "tool_protocol_adapter.bridge.lora_magnitude_vector",
     "toolspec_bridge",
     "toolspec_bridge.adapter",
-    "toolspec_bridge.adapter.base_layer",
-    "toolspec_bridge.adapter.lora_A",
-    "toolspec_bridge.adapter.lora_A.default",
-    "toolspec_bridge.adapter.lora_B",
-    "toolspec_bridge.adapter.lora_B.default",
-    "toolspec_bridge.adapter.lora_dropout",
-    "toolspec_bridge.adapter.lora_dropout.default",
-    "toolspec_bridge.adapter.lora_embedding_A",
-    "toolspec_bridge.adapter.lora_embedding_B",
-    "toolspec_bridge.adapter.lora_magnitude_vector",
     "toolspec_head",
     "toolspec_head.param_schema_adapter",
-    "toolspec_head.param_schema_adapter.base_layer",
-    "toolspec_head.param_schema_adapter.lora_A",
-    "toolspec_head.param_schema_adapter.lora_A.default",
-    "toolspec_head.param_schema_adapter.lora_B",
-    "toolspec_head.param_schema_adapter.lora_B.default",
-    "toolspec_head.param_schema_adapter.lora_dropout",
-    "toolspec_head.param_schema_adapter.lora_dropout.default",
-    "toolspec_head.param_schema_adapter.lora_embedding_A",
-    "toolspec_head.param_schema_adapter.lora_embedding_B",
-    "toolspec_head.param_schema_adapter.lora_magnitude_vector",
     "toolspec_head.postcondition_adapter",
-    "toolspec_head.postcondition_adapter.base_layer",
-    "toolspec_head.postcondition_adapter.lora_A",
-    "toolspec_head.postcondition_adapter.lora_A.default",
-    "toolspec_head.postcondition_adapter.lora_B",
-    "toolspec_head.postcondition_adapter.lora_B.default",
-    "toolspec_head.postcondition_adapter.lora_dropout",
-    "toolspec_head.postcondition_adapter.lora_dropout.default",
-    "toolspec_head.postcondition_adapter.lora_embedding_A",
-    "toolspec_head.postcondition_adapter.lora_embedding_B",
-    "toolspec_head.postcondition_adapter.lora_magnitude_vector",
     "toolspec_head.precondition_adapter",
-    "toolspec_head.precondition_adapter.base_layer",
-    "toolspec_head.precondition_adapter.lora_A",
-    "toolspec_head.precondition_adapter.lora_A.default",
-    "toolspec_head.precondition_adapter.lora_B",
-    "toolspec_head.precondition_adapter.lora_B.default",
-    "toolspec_head.precondition_adapter.lora_dropout",
-    "toolspec_head.precondition_adapter.lora_dropout.default",
-    "toolspec_head.precondition_adapter.lora_embedding_A",
-    "toolspec_head.precondition_adapter.lora_embedding_B",
-    "toolspec_head.precondition_adapter.lora_magnitude_vector",
     "toolspec_head.receipt_formatter",
-    "toolspec_head.receipt_formatter.base_layer",
-    "toolspec_head.receipt_formatter.lora_A",
-    "toolspec_head.receipt_formatter.lora_A.default",
-    "toolspec_head.receipt_formatter.lora_B",
-    "toolspec_head.receipt_formatter.lora_B.default",
-    "toolspec_head.receipt_formatter.lora_dropout",
-    "toolspec_head.receipt_formatter.lora_dropout.default",
-    "toolspec_head.receipt_formatter.lora_embedding_A",
-    "toolspec_head.receipt_formatter.lora_embedding_B",
-    "toolspec_head.receipt_formatter.lora_magnitude_vector",
     "toolspec_head.trigger_gate",
-    "toolspec_head.trigger_gate.base_layer",
-    "toolspec_head.trigger_gate.lora_A",
-    "toolspec_head.trigger_gate.lora_A.default",
-    "toolspec_head.trigger_gate.lora_B",
-    "toolspec_head.trigger_gate.lora_B.default",
-    "toolspec_head.trigger_gate.lora_dropout",
-    "toolspec_head.trigger_gate.lora_dropout.default",
-    "toolspec_head.trigger_gate.lora_embedding_A",
-    "toolspec_head.trigger_gate.lora_embedding_B",
-    "toolspec_head.trigger_gate.lora_magnitude_vector",
     "toolspec_head.validator_gate",
-    "toolspec_head.validator_gate.base_layer",
-    "toolspec_head.validator_gate.lora_A",
-    "toolspec_head.validator_gate.lora_A.default",
-    "toolspec_head.validator_gate.lora_B",
-    "toolspec_head.validator_gate.lora_B.default",
-    "toolspec_head.validator_gate.lora_dropout",
-    "toolspec_head.validator_gate.lora_dropout.default",
-    "toolspec_head.validator_gate.lora_embedding_A",
-    "toolspec_head.validator_gate.lora_embedding_B",
-    "toolspec_head.validator_gate.lora_magnitude_vector",
     "uncertainty_head",
     "uncertainty_head.calibration_mlp",
     "uncertainty_head.proj",
     "validator_feedback_bridge",
     "validator_feedback_bridge.adapter",
-    "validator_feedback_bridge.adapter.base_layer",
-    "validator_feedback_bridge.adapter.lora_A",
-    "validator_feedback_bridge.adapter.lora_A.default",
-    "validator_feedback_bridge.adapter.lora_B",
-    "validator_feedback_bridge.adapter.lora_B.default",
-    "validator_feedback_bridge.adapter.lora_dropout",
-    "validator_feedback_bridge.adapter.lora_dropout.default",
-    "validator_feedback_bridge.adapter.lora_embedding_A",
-    "validator_feedback_bridge.adapter.lora_embedding_B",
-    "validator_feedback_bridge.adapter.lora_magnitude_vector",
     "validator_feedback_loop",
     "validator_feedback_loop.rollback_adapter",
     "validator_receipt_bridge",
@@ -228,10 +138,10 @@
     "backbone_frozen": true,
     "backbone_present": true,
     "hidden_size": 2048,
-    "host_parameters": 176621573,
-    "host_trainable_parameters": 360448,
-    "total_parameters": 34132679301,
-    "trainable_parameters": 360448,
     "version": "veriloop.coder_peft_host.v1"
   },
   "version": "veriloop.coder_peft_host.v1"

     "query_runtime_engine.adapter",
     "request_normalizer",
     "request_normalizer.adapter",
     "rollback_adapter",
     "rollback_adapter.head",
     "rollback_engine",
     "runtime_binding.adapter",
     "runtime_harness_adapter",
     "runtime_harness_adapter.bridge",
     "runtime_harness_uncertainty_bridge",
     "runtime_harness_uncertainty_bridge.adapter",
     "sandbox_rollback_bridge",
     "session_state_manager.rollback_state",
     "tool_protocol_adapter",
     "tool_protocol_adapter.bridge",
     "toolspec_bridge",
     "toolspec_bridge.adapter",
     "toolspec_head",
     "toolspec_head.param_schema_adapter",
     "toolspec_head.postcondition_adapter",
     "toolspec_head.precondition_adapter",
     "toolspec_head.receipt_formatter",
     "toolspec_head.trigger_gate",
     "toolspec_head.validator_gate",
     "uncertainty_head",
     "uncertainty_head.calibration_mlp",
+    "uncertainty_head.calibration_mlp.base_layer",
+    "uncertainty_head.calibration_mlp.lora_A",
+    "uncertainty_head.calibration_mlp.lora_A.default",
+    "uncertainty_head.calibration_mlp.lora_B",
+    "uncertainty_head.calibration_mlp.lora_B.default",
+    "uncertainty_head.calibration_mlp.lora_dropout",
+    "uncertainty_head.calibration_mlp.lora_dropout.default",
+    "uncertainty_head.calibration_mlp.lora_embedding_A",
+    "uncertainty_head.calibration_mlp.lora_embedding_B",
+    "uncertainty_head.calibration_mlp.lora_magnitude_vector",
     "uncertainty_head.proj",
+    "uncertainty_head.proj.base_layer",
+    "uncertainty_head.proj.lora_A",
+    "uncertainty_head.proj.lora_A.default",
+    "uncertainty_head.proj.lora_B",
+    "uncertainty_head.proj.lora_B.default",
+    "uncertainty_head.proj.lora_dropout",
+    "uncertainty_head.proj.lora_dropout.default",
+    "uncertainty_head.proj.lora_embedding_A",
+    "uncertainty_head.proj.lora_embedding_B",
+    "uncertainty_head.proj.lora_magnitude_vector",
     "validator_feedback_bridge",
     "validator_feedback_bridge.adapter",
     "validator_feedback_loop",
     "validator_feedback_loop.rollback_adapter",
     "validator_receipt_bridge",
     "backbone_frozen": true,
     "backbone_present": true,
     "hidden_size": 2048,
+    "host_parameters": 176310317,
+    "host_trainable_parameters": 49192,
+    "total_parameters": 34132368045,
+    "trainable_parameters": 49192,
     "version": "veriloop.coder_peft_host.v1"
   },
   "version": "veriloop.coder_peft_host.v1"

tokenizer/tokenizer.json CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a0316e38d81a31965b770393461dcc03612ac06e2733f8e1cc440d36a2ef189
 size 19989442

 version https://git-lfs.github.com/spec/v1
+oid sha256:ea2e66b594a0906e9a547c9e6ff9e5fb8a8198439c8cf7d6dc48f23529161223
 size 19989442

uncertainty_adapter_plan.json ADDED Viewed

	@@ -0,0 +1,469 @@

+{
+  "adapter_alpha": 16,
+  "adapter_dropout": 0.05,
+  "adapter_rank": 8,
+  "backbone_manifest": {
+    "host_inventory_size": 79
+  },
+  "config": {
+    "allow_backbone_bridge": false,
+    "backbone": "/public/wang_libo/veriloop_coder_e1/model",
+    "bf16": true,
+    "cache_dir": null,
+    "cpu_max_memory_gib": 96,
+    "dataset_jsonl": null,
+    "dim_weights": {
+      "u_answer": 1.1,
+      "u_evidence": 1.35,
+      "u_exec": 1.35,
+      "u_risk": 1.55,
+      "u_spec": 1.25
+    },
+    "early_stopping_min_delta": 0.001,
+    "early_stopping_patience": 2,
+    "enable_synthetic_dataset": true,
+    "eval_jsonl": null,
+    "eval_samples_per_mode": 5,
+    "fp16": false,
+    "gpu_max_memory_gib": 44,
+    "gradient_accumulation_steps": 16,
+    "host_dropout": 0.0,
+    "learning_rate": 2e-05,
+    "local_files_only": true,
+    "logging_steps": 10,
+    "max_grad_norm": 1.0,
+    "max_length": 1664,
+    "min_epochs_before_early_stop": 2,
+    "num_train_epochs": 4.0,
+    "output_dir": "./outputs/uncertainty_qwen36_rootfix_run1",
+    "per_device_eval_batch_size": 1,
+    "per_device_train_batch_size": 1,
+    "prefer_best_checkpoint_export": true,
+    "probe_dropout": 0.03,
+    "probe_hidden_factor": 0.75,
+    "product_line": "veriloop_coder",
+    "quantization_mode": "4bit",
+    "quantization_required": false,
+    "require_trainable_targets": true,
+    "revision": null,
+    "run_post_train_eval": true,
+    "save_best_checkpoint": true,
+    "seed": 11,
+    "selection_mode": "minimal",
+    "train_samples_per_mode": 14,
+    "training_mode": "mounted_head",
+    "trust_remote_code": true,
+    "use_double_quant": true,
+    "warmup_ratio": 0.05,
+    "weight_decay": 0.0,
+    "weighted_mae_penalty": 0.5,
+    "weighted_rmse_penalty": 0.5
+  },
+  "dataset_summary": {
+    "eval_modes": {
+      "conflicting_evidence": 5,
+      "evidence_gap": 5,
+      "exec_required": 5,
+      "high_risk": 5,
+      "low_uncertainty": 5,
+      "patch_pending": 5,
+      "reverse_engineering_ambiguity": 5,
+      "self_check_failure": 5,
+      "spec_mismatch": 5,
+      "validator_negation": 5,
+      "worktree_conflict": 5
+    },
+    "eval_size": 55,
+    "train_modes": {
+      "conflicting_evidence": 14,
+      "evidence_gap": 14,
+      "exec_required": 14,
+      "high_risk": 14,
+      "low_uncertainty": 14,
+      "patch_pending": 14,
+      "reverse_engineering_ambiguity": 14,
+      "self_check_failure": 14,
+      "spec_mismatch": 14,
+      "validator_negation": 14,
+      "worktree_conflict": 14
+    },
+    "train_size": 154
+  },
+  "head_strategy": "host_head",
+  "notes": [
+    "Primary route is host-surface-first uncertainty training.",
+    "Validator receipts, self-check failure, reverse-engineering ambiguity, worktree conflicts, and patch continuity are first-class signals.",
+    "DualPath, Full AttnRes, mHC, routers, experts, and broad attention-layer PEFT stay excluded.",
+    "This adapter should improve runtime uncertainty routing, not general coding free-formity."
+  ],
+  "peft_method": "lora_narrow",
+  "recipe": {
+    "adapter_family": "uncertainty",
+    "backbone": "/public/wang_libo/veriloop_coder_e1/model",
+    "backbone_family": "qwen_dense",
+    "excluded_patterns": [
+      "(?i)\\bdualpath\\b",
+      "(?i)\\bmhc\\b",
+      "(?i)\\bfull[_\\- ]?attnres\\b",
+      "(?i)\\battnres(_full)?\\b",
+      "(?i)\\brouter\\b",
+      "(?i)\\bexperts?\\b",
+      "(?i)\\bmoe\\b.*\\b(gate|router|expert)\\b",
+      "(?i)\\brope\\b.*\\b(freq|inv_freq|theta|rotary)\\b",
+      "(?i)\\bkvcache\\b",
+      "(?i)\\bposition_embedding\\b",
+      "(?i)\\bembed(tokens|ding)?\\b",
+      "(?i)\\blm_head\\b"
+    ],
+    "harness_constraints": [
+      "Harness Engineering remains the primary convergence layer.",
+      "Adapter must not bypass runtime orchestrator / validator / rollback loops.",
+      "Adapter outputs remain subordinate to VeriLoop control-plane decisions.",
+      "Adapter must not create hidden prompt-style memory authority.",
+      "Adapter must support bounded uncertainty calibration rather than generic hesitation.",
+      "Validator and receipt evidence must remain able to update uncertainty."
+    ],
+    "hyperparams": {
+      "alpha": 16,
+      "bias": "none",
+      "dropout": 0.05,
+      "fan_in_fan_out": false,
+      "modules_to_save": [
+        "input_layernorm"
+      ],
+      "r": 8,
+      "task_type": "CAUSAL_LM"
+    },
+    "merge_policy": "merge_after_guard",
+    "metadata": {
+      "allow_backbone_bridge": false,
+      "allow_vla_action_expert": false,
+      "harness_first": true,
+      "prefer_explicit_heads": true,
+      "prefer_qlora_for_backbone_bridge": true,
+      "require_harness_first": true,
+      "selector_group_count": 2,
+      "strict_narrow_scope": true,
+      "trainer": "veriloop.uncertainty_adapter_trainer.v5.qwen36",
+      "uncertainty_training": true
+    },
+    "notes": [
+      "Backbone bridge tuning disabled explicitly; selector stays on custom surfaces or no-op.",
+      "Backbone family inferred as qwen_dense.",
+      "PEFT method resolved as lora_narrow.",
+      "Recipe is harness-first: runtime convergence remains in VeriLoop control-plane + harness, not in broad weight surgery.",
+      "Block AttnRes, DualPath, mHC hooks, RoPE, KV-cache, and broad MoE routing remain structurally excluded."
+    ],
+    "peft_method": "lora_narrow",
+    "precision_policy": "auto",
+    "product_line": "veriloop_coder",
+    "regression_requirements": [
+      "Must pass PEFT regression guard structural policy checks.",
+      "Must not introduce forbidden backbone/serving structural targets.",
+      "Must preserve harness regression envelope for the selected product line.",
+      "Budgeted uncertainty convergence must not regress.",
+      "Uncertainty calibration must not collapse into generic caution."
+    ],
+    "target_groups": [
+      {
+        "alpha": 16,
+        "dropout": 0.0,
+        "name": "group_1_custom_control_head",
+        "rank": 8,
+        "rationale": "Prefer explicit uncertainty / calibration heads over backbone surgery.",
+        "surface": "custom_control_head",
+        "target_modules": [
+          "uncertainty_head",
+          "uncertainty_head.calibration_mlp",
+          "uncertainty_head.proj"
+        ]
+      },
+      {
+        "alpha": 16,
+        "dropout": 0.0,
+        "name": "group_2_custom_validator_bridge",
+        "rank": 8,
+        "rationale": "Validation and rollback fidelity should prefer explicit validator / rollback bridges.",
+        "surface": "custom_validator_bridge",
+        "target_modules": [
+          "failure_signal_bridge",
+          "failure_signal_bridge.rollback_bridge",
+          "rollback_adapter",
+          "rollback_adapter.head",
+          "sandbox_rollback_bridge",
+          "sandbox_rollback_bridge.adapter",
+          "validator_feedback_bridge",
+          "validator_feedback_bridge.adapter",
+          "validator_feedback_loop.rollback_adapter"
+        ]
+      }
+    ],
+    "target_modules": [
+      "uncertainty_head",
+      "uncertainty_head.calibration_mlp",
+      "uncertainty_head.proj",
+      "failure_signal_bridge",
+      "failure_signal_bridge.rollback_bridge",
+      "rollback_adapter",
+      "rollback_adapter.head",
+      "sandbox_rollback_bridge",
+      "sandbox_rollback_bridge.adapter",
+      "validator_feedback_bridge",
+      "validator_feedback_bridge.adapter",
+      "validator_feedback_loop.rollback_adapter"
+    ],
+    "version": "veriloop.lora_recipe_veriloop.v2"
+  },
+  "selected_target_modules": [
+    "uncertainty_head",
+    "uncertainty_head.calibration_mlp",
+    "uncertainty_head.proj"
+  ],
+  "target_selection": {
+    "backbone_archetype": "qwen_dense",
+    "exclusions": [
+      {
+        "pattern": "(^|\\.)lm_head($|\\.)",
+        "reason": "Do not retune final token head; too broad and evaluation-heavy."
+      },
+      {
+        "pattern": "(^|\\.)embed_tokens($|\\.)",
+        "reason": "Embedding surgery risks broad semantic drift."
+      },
+      {
+        "pattern": "(^|\\.)norm($|\\.)",
+        "reason": "Global norm tuning can destabilize calibration across scenes."
+      },
+      {
+        "pattern": "attnres|attention_residual",
+        "reason": "Block AttnRes may be mounted structurally but is never a PEFT target."
+      },
+      {
+        "pattern": "dualpath",
+        "reason": "DualPath is serving/runtime infrastructure only."
+      },
+      {
+        "pattern": "mhc|hyper[-_]?connection",
+        "reason": "mHC-inspired stability hooks remain structural, not PEFT surfaces."
+      },
+      {
+        "pattern": "rope|rotary",
+        "reason": "RoPE/context surgery is handled architecturally, not by narrow PEFT here."
+      },
+      {
+        "pattern": "kvcache|kv_cache",
+        "reason": "KV-cache runtime surfaces are not PEFT targets."
+      },
+      {
+        "pattern": "(^|\\.)memory(_store|_bank)?($|\\.)",
+        "reason": "Persistent memory stores are harness/runtime policy surfaces, not PEFT targets."
+      }
+    ],
+    "inventory_size": 79,
+    "inventory_source": "provided_names",
+    "notes": [
+      "Harness Engineering is primary; PEFT is limited to obedience-facing, interface-facing support surfaces.",
+      "Backbone bridge tuning disabled explicitly; selector stays on custom surfaces or no-op."
+    ],
+    "recommended_training": "ia3_head_only",
+    "request": {
+      "allow_backbone_bridge": false,
+      "allow_vla_action_expert": false,
+      "backbone": "/public/wang_libo/veriloop_coder_e1/model",
+      "explicit_exclude_patterns": [],
+      "explicit_include_patterns": [],
+      "intents": [
+        "uncertainty",
+        "validator_alignment",
+        "harness_alignment",
+        "runtime_protocol",
+        "session_continuity",
+        "worktree_discipline"
+      ],
+      "prefer_qlora_for_backbone_bridge": true,
+      "product_line": "veriloop_coder",
+      "selection_mode": "minimal"
+    },
+    "selected_groups": [
+      {
+        "alpha": 16,
+        "dropout": 0.0,
+        "exclude_patterns": [
+          "(^|\\.)lm_head($|\\.)",
+          "(^|\\.)embed_tokens($|\\.)",
+          "(^|\\.)norm($|\\.)",
+          "attnres|attention_residual",
+          "dualpath",
+          "mhc|hyper[-_]?connection",
+          "rope|rotary",
+          "kvcache|kv_cache",
+          "(^|\\.)memory(_store|_bank)?($|\\.)"
+        ],
+        "include_patterns": [
+          "(^|\\.)(uncertainty_head|uncertainty_adapter|calib(_head|ration_adapter)?)($|\\.)"
+        ],
+        "intents": [
+          "uncertainty"
+        ],
+        "layer_window": {
+          "mode": "all",
+          "value": 0.0
+        },
+        "matched_module_names": [
+          "uncertainty_head",
+          "uncertainty_head.calibration_mlp",
+          "uncertainty_head.proj"
+        ],
+        "name": "group_1_custom_control_head",
+        "rank": 8,
+        "rationale": "Prefer explicit uncertainty / calibration heads over backbone surgery.",
+        "risk": "low",
+        "surface": "custom_control_head"
+      },
+      {
+        "alpha": 16,
+        "dropout": 0.05,
+        "exclude_patterns": [
+          "(^|\\.)lm_head($|\\.)",
+          "(^|\\.)embed_tokens($|\\.)",
+          "(^|\\.)norm($|\\.)",
+          "attnres|attention_residual",
+          "dualpath",
+          "mhc|hyper[-_]?connection",
+          "rope|rotary",
+          "kvcache|kv_cache",
+          "(^|\\.)memory(_store|_bank)?($|\\.)"
+        ],
+        "include_patterns": [
+          "(^|\\.)(runtime_harness|query_runtime|task_brief|task_runtime|tool_protocol|permission_context|worktree|session_state|request_normalizer|action_allowlist|constraint_guard|progress_state|workspace_snapshot|repo_contract|knowledge_entry|completion_criteria|search_bridge|sandbox_search_bridge)(_adapter|_bridge|_head)?($|\\.)",
+          "(^|\\.)(toolspec|tool(_call)?(_grammar|_interface)?|harness|validator|rollback|receipt|patch|permission|session|worktree)(_adapter|_bridge|_head)?($|\\.)"
+        ],
+        "intents": [
+          "harness_alignment",
+          "runtime_protocol",
+          "session_continuity",
+          "worktree_discipline"
+        ],
+        "layer_window": {
+          "mode": "all",
+          "value": 0.0
+        },
+        "matched_module_names": [
+          "failure_signal_bridge.rollback_bridge",
+          "request_normalizer",
+          "request_normalizer.adapter",
+          "rollback_adapter",
+          "rollback_adapter.head",
+          "runtime_harness_adapter",
+          "runtime_harness_adapter.bridge",
+          "tool_protocol_adapter",
+          "tool_protocol_adapter.bridge",
+          "toolspec_bridge",
+          "toolspec_bridge.adapter",
+          "toolspec_head",
+          "toolspec_head.param_schema_adapter",
+          "toolspec_head.postcondition_adapter",
+          "toolspec_head.precondition_adapter",
+          "toolspec_head.receipt_formatter",
+          "toolspec_head.trigger_gate",
+          "toolspec_head.validator_gate",
+          "validator_feedback_loop.rollback_adapter"
+        ],
+        "name": "group_2_custom_runtime_harness_bridge",
+        "rank": 8,
+        "rationale": "Runtime / harness obedience should attach to explicit interface bridges before any backbone fallback.",
+        "risk": "low",
+        "surface": "custom_runtime_harness_bridge"
+      },
+      {
+        "alpha": 16,
+        "dropout": 0.0,
+        "exclude_patterns": [
+          "(^|\\.)lm_head($|\\.)",
+          "(^|\\.)embed_tokens($|\\.)",
+          "(^|\\.)norm($|\\.)",
+          "attnres|attention_residual",
+          "dualpath",
+          "mhc|hyper[-_]?connection",
+          "rope|rotary",
+          "kvcache|kv_cache",
+          "(^|\\.)memory(_store|_bank)?($|\\.)"
+        ],
+        "include_patterns": [
+          "(^|\\.)(validator_feedback|sandbox_result_validator|sandbox_rollback_bridge|failure_signal|rollback)(_adapter|_bridge|_head)?($|\\.)"
+        ],
+        "intents": [
+          "validator_alignment"
+        ],
+        "layer_window": {
+          "mode": "all",
+          "value": 0.0
+        },
+        "matched_module_names": [
+          "failure_signal_bridge",
+          "failure_signal_bridge.rollback_bridge",
+          "rollback_adapter",
+          "rollback_adapter.head",
+          "sandbox_rollback_bridge",
+          "sandbox_rollback_bridge.adapter",
+          "validator_feedback_bridge",
+          "validator_feedback_bridge.adapter",
+          "validator_feedback_loop.rollback_adapter"
+        ],
+        "name": "group_3_custom_validator_bridge",
+        "rank": 8,
+        "rationale": "Validation and rollback fidelity should prefer explicit validator / rollback bridges.",
+        "risk": "low",
+        "surface": "custom_validator_bridge"
+      },
+      {
+        "alpha": 8,
+        "dropout": 0.0,
+        "exclude_patterns": [
+          "(^|\\.)lm_head($|\\.)",
+          "(^|\\.)embed_tokens($|\\.)",
+          "(^|\\.)norm($|\\.)",
+          "attnres|attention_residual",
+          "dualpath",
+          "mhc|hyper[-_]?connection",
+          "rope|rotary",
+          "kvcache|kv_cache",
+          "(^|\\.)memory(_store|_bank)?($|\\.)"
+        ],
+        "include_patterns": [
+          "(^|\\.)(memory_boundary_guard|episodic_memory|session_compactor)(_adapter|_bridge|_head)?($|\\.)"
+        ],
+        "intents": [
+          "session_continuity"
+        ],
+        "layer_window": {
+          "mode": "all",
+          "value": 0.0
+        },
+        "matched_module_names": [
+          "episodic_memory",
+          "episodic_memory.adapter",
+          "memory_boundary_guard",
+          "memory_boundary_guard.adapter",
+          "memory_boundary_guard.rollback_filter",
+          "session_compactor",
+          "session_compactor.adapter"
+        ],
+        "name": "group_4_custom_memory_boundary_bridge",
+        "rank": 4,
+        "rationale": "Session continuity should bind to boundary-aware memory packet surfaces rather than broad backbone tuning.",
+        "risk": "low",
+        "surface": "custom_memory_boundary_bridge"
+      }
+    ],
+    "upstream_profile_hint": null,
+    "version": "veriloop.peft_target_selector.v2",
+    "warnings": []
+  },
+  "training_mode": "mounted_head",
+  "version": "veriloop.uncertainty_adapter_trainer.v5.qwen36",
+  "warnings": [
+    "Harness Engineering is primary; PEFT is limited to obedience-facing, interface-facing support surfaces.",
+    "Backbone bridge tuning disabled explicitly; selector stays on custom surfaces or no-op."
+  ]
+}

uncertainty_adapter_train_result.json ADDED Viewed

	@@ -0,0 +1,564 @@

+{
+  "artifacts": {
+    "adapter_dir": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/adapter",
+    "best_checkpoint_manifest": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/best_checkpoint_manifest.json",
+    "epoch_history": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/epoch_history.json",
+    "eval_jsonl": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/uncertainty_eval.jsonl",
+    "host_manifest": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/host_manifest.json",
+    "plan_json": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/uncertainty_adapter_plan.json",
+    "tokenizer_dir": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/tokenizer",
+    "train_jsonl": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/uncertainty_train.jsonl",
+    "training_manifest": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/uncertainty_training_manifest.json",
+    "uncertainty_head": "/private/wang_libo/veriloop_coder_e1/outputs/uncertainty_qwen36_rootfix_run1/uncertainty_head.pt"
+  },
+  "dataset": {
+    "eval_modes": {
+      "conflicting_evidence": 5,
+      "evidence_gap": 5,
+      "exec_required": 5,
+      "high_risk": 5,
+      "low_uncertainty": 5,
+      "patch_pending": 5,
+      "reverse_engineering_ambiguity": 5,
+      "self_check_failure": 5,
+      "spec_mismatch": 5,
+      "validator_negation": 5,
+      "worktree_conflict": 5
+    },
+    "eval_size": 55,
+    "train_modes": {
+      "conflicting_evidence": 14,
+      "evidence_gap": 14,
+      "exec_required": 14,
+      "high_risk": 14,
+      "low_uncertainty": 14,
+      "patch_pending": 14,
+      "reverse_engineering_ambiguity": 14,
+      "self_check_failure": 14,
+      "spec_mismatch": 14,
+      "validator_negation": 14,
+      "worktree_conflict": 14
+    },
+    "train_size": 154
+  },
+  "eval_metrics": {
+    "adapter_exported": true,
+    "auto_lora_from_ia3": false,
+    "best_epoch": 2,
+    "best_quality_score": 0.8625935807221907,
+    "count": 55,
+    "mae": {
+      "u_answer": 0.15174226462841034,
+      "u_evidence": 0.19610758125782013,
+      "u_exec": 0.18561214208602905,
+      "u_risk": 0.1553734689950943,
+      "u_spec": 0.21633382141590118
+    },
+    "mean_mae": 0.18103384971618652,
+    "mean_rmse": 0.24169571697711945,
+    "moderate_accuracy": 0.6727272727272727,
+    "peft_method": "lora_narrow",
+    "quality_score": 0.8625935807221907,
+    "rmse": {
+      "u_answer": 0.18724055588245392,
+      "u_evidence": 0.22527915239334106,
+      "u_exec": 0.25238174200057983,
+      "u_risk": 0.20667441189289093,
+      "u_spec": 0.3369026482105255
+    },
+    "tight_accuracy": 0.4,
+    "used_peft": true,
+    "weighted_mae": 0.18083095811830807,
+    "weighted_rmse": 0.24125460771003793
+  },
+  "plan": {
+    "adapter_alpha": 16,
+    "adapter_dropout": 0.05,
+    "adapter_rank": 8,
+    "backbone_manifest": {
+      "host_inventory_size": 79
+    },
+    "config": {
+      "allow_backbone_bridge": false,
+      "backbone": "/public/wang_libo/veriloop_coder_e1/model",
+      "bf16": true,
+      "cache_dir": null,
+      "cpu_max_memory_gib": 96,
+      "dataset_jsonl": null,
+      "dim_weights": {
+        "u_answer": 1.1,
+        "u_evidence": 1.35,
+        "u_exec": 1.35,
+        "u_risk": 1.55,
+        "u_spec": 1.25
+      },
+      "early_stopping_min_delta": 0.001,
+      "early_stopping_patience": 2,
+      "enable_synthetic_dataset": true,
+      "eval_jsonl": null,
+      "eval_samples_per_mode": 5,
+      "fp16": false,
+      "gpu_max_memory_gib": 44,
+      "gradient_accumulation_steps": 16,
+      "host_dropout": 0.0,
+      "learning_rate": 2e-05,
+      "local_files_only": true,
+      "logging_steps": 10,
+      "max_grad_norm": 1.0,
+      "max_length": 1664,
+      "min_epochs_before_early_stop": 2,
+      "num_train_epochs": 4.0,
+      "output_dir": "./outputs/uncertainty_qwen36_rootfix_run1",
+      "per_device_eval_batch_size": 1,
+      "per_device_train_batch_size": 1,
+      "prefer_best_checkpoint_export": true,
+      "probe_dropout": 0.03,
+      "probe_hidden_factor": 0.75,
+      "product_line": "veriloop_coder",
+      "quantization_mode": "4bit",
+      "quantization_required": false,
+      "require_trainable_targets": true,
+      "revision": null,
+      "run_post_train_eval": true,
+      "save_best_checkpoint": true,
+      "seed": 11,
+      "selection_mode": "minimal",
+      "train_samples_per_mode": 14,
+      "training_mode": "mounted_head",
+      "trust_remote_code": true,
+      "use_double_quant": true,
+      "warmup_ratio": 0.05,
+      "weight_decay": 0.0,
+      "weighted_mae_penalty": 0.5,
+      "weighted_rmse_penalty": 0.5
+    },
+    "dataset_summary": {
+      "eval_modes": {
+        "conflicting_evidence": 5,
+        "evidence_gap": 5,
+        "exec_required": 5,
+        "high_risk": 5,
+        "low_uncertainty": 5,
+        "patch_pending": 5,
+        "reverse_engineering_ambiguity": 5,
+        "self_check_failure": 5,
+        "spec_mismatch": 5,
+        "validator_negation": 5,
+        "worktree_conflict": 5
+      },
+      "eval_size": 55,
+      "train_modes": {
+        "conflicting_evidence": 14,
+        "evidence_gap": 14,
+        "exec_required": 14,
+        "high_risk": 14,
+        "low_uncertainty": 14,
+        "patch_pending": 14,
+        "reverse_engineering_ambiguity": 14,
+        "self_check_failure": 14,
+        "spec_mismatch": 14,
+        "validator_negation": 14,
+        "worktree_conflict": 14
+      },
+      "train_size": 154
+    },
+    "head_strategy": "host_head",
+    "notes": [
+      "Primary route is host-surface-first uncertainty training.",
+      "Validator receipts, self-check failure, reverse-engineering ambiguity, worktree conflicts, and patch continuity are first-class signals.",
+      "DualPath, Full AttnRes, mHC, routers, experts, and broad attention-layer PEFT stay excluded.",
+      "This adapter should improve runtime uncertainty routing, not general coding free-formity."
+    ],
+    "peft_method": "lora_narrow",
+    "recipe": {
+      "adapter_family": "uncertainty",
+      "backbone": "/public/wang_libo/veriloop_coder_e1/model",
+      "backbone_family": "qwen_dense",
+      "excluded_patterns": [
+        "(?i)\\bdualpath\\b",
+        "(?i)\\bmhc\\b",
+        "(?i)\\bfull[_\\- ]?attnres\\b",
+        "(?i)\\battnres(_full)?\\b",
+        "(?i)\\brouter\\b",
+        "(?i)\\bexperts?\\b",
+        "(?i)\\bmoe\\b.*\\b(gate|router|expert)\\b",
+        "(?i)\\brope\\b.*\\b(freq|inv_freq|theta|rotary)\\b",
+        "(?i)\\bkvcache\\b",
+        "(?i)\\bposition_embedding\\b",
+        "(?i)\\bembed(tokens|ding)?\\b",
+        "(?i)\\blm_head\\b"
+      ],
+      "harness_constraints": [
+        "Harness Engineering remains the primary convergence layer.",
+        "Adapter must not bypass runtime orchestrator / validator / rollback loops.",
+        "Adapter outputs remain subordinate to VeriLoop control-plane decisions.",
+        "Adapter must not create hidden prompt-style memory authority.",
+        "Adapter must support bounded uncertainty calibration rather than generic hesitation.",
+        "Validator and receipt evidence must remain able to update uncertainty."
+      ],
+      "hyperparams": {
+        "alpha": 16,
+        "bias": "none",
+        "dropout": 0.05,
+        "fan_in_fan_out": false,
+        "modules_to_save": [
+          "input_layernorm"
+        ],
+        "r": 8,
+        "task_type": "CAUSAL_LM"
+      },
+      "merge_policy": "merge_after_guard",
+      "metadata": {
+        "allow_backbone_bridge": false,
+        "allow_vla_action_expert": false,
+        "harness_first": true,
+        "prefer_explicit_heads": true,
+        "prefer_qlora_for_backbone_bridge": true,
+        "require_harness_first": true,
+        "selector_group_count": 2,
+        "strict_narrow_scope": true,
+        "trainer": "veriloop.uncertainty_adapter_trainer.v5.qwen36",
+        "uncertainty_training": true
+      },
+      "notes": [
+        "Backbone bridge tuning disabled explicitly; selector stays on custom surfaces or no-op.",
+        "Backbone family inferred as qwen_dense.",
+        "PEFT method resolved as lora_narrow.",
+        "Recipe is harness-first: runtime convergence remains in VeriLoop control-plane + harness, not in broad weight surgery.",
+        "Block AttnRes, DualPath, mHC hooks, RoPE, KV-cache, and broad MoE routing remain structurally excluded."
+      ],
+      "peft_method": "lora_narrow",
+      "precision_policy": "auto",
+      "product_line": "veriloop_coder",
+      "regression_requirements": [
+        "Must pass PEFT regression guard structural policy checks.",
+        "Must not introduce forbidden backbone/serving structural targets.",
+        "Must preserve harness regression envelope for the selected product line.",
+        "Budgeted uncertainty convergence must not regress.",
+        "Uncertainty calibration must not collapse into generic caution."
+      ],
+      "target_groups": [
+        {
+          "alpha": 16,
+          "dropout": 0.0,
+          "name": "group_1_custom_control_head",
+          "rank": 8,
+          "rationale": "Prefer explicit uncertainty / calibration heads over backbone surgery.",
+          "surface": "custom_control_head",
+          "target_modules": [
+            "uncertainty_head",
+            "uncertainty_head.calibration_mlp",
+            "uncertainty_head.proj"
+          ]
+        },
+        {
+          "alpha": 16,
+          "dropout": 0.0,
+          "name": "group_2_custom_validator_bridge",
+          "rank": 8,
+          "rationale": "Validation and rollback fidelity should prefer explicit validator / rollback bridges.",
+          "surface": "custom_validator_bridge",
+          "target_modules": [
+            "failure_signal_bridge",
+            "failure_signal_bridge.rollback_bridge",
+            "rollback_adapter",
+            "rollback_adapter.head",
+            "sandbox_rollback_bridge",
+            "sandbox_rollback_bridge.adapter",
+            "validator_feedback_bridge",
+            "validator_feedback_bridge.adapter",
+            "validator_feedback_loop.rollback_adapter"
+          ]
+        }
+      ],
+      "target_modules": [
+        "uncertainty_head",
+        "uncertainty_head.calibration_mlp",
+        "uncertainty_head.proj",
+        "failure_signal_bridge",
+        "failure_signal_bridge.rollback_bridge",
+        "rollback_adapter",
+        "rollback_adapter.head",
+        "sandbox_rollback_bridge",
+        "sandbox_rollback_bridge.adapter",
+        "validator_feedback_bridge",
+        "validator_feedback_bridge.adapter",
+        "validator_feedback_loop.rollback_adapter"
+      ],
+      "version": "veriloop.lora_recipe_veriloop.v2"
+    },
+    "selected_target_modules": [
+      "uncertainty_head",
+      "uncertainty_head.calibration_mlp",
+      "uncertainty_head.proj"
+    ],
+    "target_selection": {
+      "backbone_archetype": "qwen_dense",
+      "exclusions": [
+        {
+          "pattern": "(^|\\.)lm_head($|\\.)",
+          "reason": "Do not retune final token head; too broad and evaluation-heavy."
+        },
+        {
+          "pattern": "(^|\\.)embed_tokens($|\\.)",
+          "reason": "Embedding surgery risks broad semantic drift."
+        },
+        {
+          "pattern": "(^|\\.)norm($|\\.)",
+          "reason": "Global norm tuning can destabilize calibration across scenes."
+        },
+        {
+          "pattern": "attnres|attention_residual",
+          "reason": "Block AttnRes may be mounted structurally but is never a PEFT target."
+        },
+        {
+          "pattern": "dualpath",
+          "reason": "DualPath is serving/runtime infrastructure only."
+        },
+        {
+          "pattern": "mhc|hyper[-_]?connection",
+          "reason": "mHC-inspired stability hooks remain structural, not PEFT surfaces."
+        },
+        {
+          "pattern": "rope|rotary",
+          "reason": "RoPE/context surgery is handled architecturally, not by narrow PEFT here."
+        },
+        {
+          "pattern": "kvcache|kv_cache",
+          "reason": "KV-cache runtime surfaces are not PEFT targets."
+        },
+        {
+          "pattern": "(^|\\.)memory(_store|_bank)?($|\\.)",
+          "reason": "Persistent memory stores are harness/runtime policy surfaces, not PEFT targets."
+        }
+      ],
+      "inventory_size": 79,
+      "inventory_source": "provided_names",
+      "notes": [
+        "Harness Engineering is primary; PEFT is limited to obedience-facing, interface-facing support surfaces.",
+        "Backbone bridge tuning disabled explicitly; selector stays on custom surfaces or no-op."
+      ],
+      "recommended_training": "ia3_head_only",
+      "request": {
+        "allow_backbone_bridge": false,
+        "allow_vla_action_expert": false,
+        "backbone": "/public/wang_libo/veriloop_coder_e1/model",
+        "explicit_exclude_patterns": [],
+        "explicit_include_patterns": [],
+        "intents": [
+          "uncertainty",
+          "validator_alignment",
+          "harness_alignment",
+          "runtime_protocol",
+          "session_continuity",
+          "worktree_discipline"
+        ],
+        "prefer_qlora_for_backbone_bridge": true,
+        "product_line": "veriloop_coder",
+        "selection_mode": "minimal"
+      },
+      "selected_groups": [
+        {
+          "alpha": 16,
+          "dropout": 0.0,
+          "exclude_patterns": [
+            "(^|\\.)lm_head($|\\.)",
+            "(^|\\.)embed_tokens($|\\.)",
+            "(^|\\.)norm($|\\.)",
+            "attnres|attention_residual",
+            "dualpath",
+            "mhc|hyper[-_]?connection",
+            "rope|rotary",
+            "kvcache|kv_cache",
+            "(^|\\.)memory(_store|_bank)?($|\\.)"
+          ],
+          "include_patterns": [
+            "(^|\\.)(uncertainty_head|uncertainty_adapter|calib(_head|ration_adapter)?)($|\\.)"
+          ],
+          "intents": [
+            "uncertainty"
+          ],
+          "layer_window": {
+            "mode": "all",
+            "value": 0.0
+          },
+          "matched_module_names": [
+            "uncertainty_head",
+            "uncertainty_head.calibration_mlp",
+            "uncertainty_head.proj"
+          ],
+          "name": "group_1_custom_control_head",
+          "rank": 8,
+          "rationale": "Prefer explicit uncertainty / calibration heads over backbone surgery.",
+          "risk": "low",
+          "surface": "custom_control_head"
+        },
+        {
+          "alpha": 16,
+          "dropout": 0.05,
+          "exclude_patterns": [
+            "(^|\\.)lm_head($|\\.)",
+            "(^|\\.)embed_tokens($|\\.)",
+            "(^|\\.)norm($|\\.)",
+            "attnres|attention_residual",
+            "dualpath",
+            "mhc|hyper[-_]?connection",
+            "rope|rotary",
+            "kvcache|kv_cache",
+            "(^|\\.)memory(_store|_bank)?($|\\.)"
+          ],
+          "include_patterns": [
+            "(^|\\.)(runtime_harness|query_runtime|task_brief|task_runtime|tool_protocol|permission_context|worktree|session_state|request_normalizer|action_allowlist|constraint_guard|progress_state|workspace_snapshot|repo_contract|knowledge_entry|completion_criteria|search_bridge|sandbox_search_bridge)(_adapter|_bridge|_head)?($|\\.)",
+            "(^|\\.)(toolspec|tool(_call)?(_grammar|_interface)?|harness|validator|rollback|receipt|patch|permission|session|worktree)(_adapter|_bridge|_head)?($|\\.)"
+          ],
+          "intents": [
+            "harness_alignment",
+            "runtime_protocol",
+            "session_continuity",
+            "worktree_discipline"
+          ],
+          "layer_window": {
+            "mode": "all",
+            "value": 0.0
+          },
+          "matched_module_names": [
+            "failure_signal_bridge.rollback_bridge",
+            "request_normalizer",
+            "request_normalizer.adapter",
+            "rollback_adapter",
+            "rollback_adapter.head",
+            "runtime_harness_adapter",
+            "runtime_harness_adapter.bridge",
+            "tool_protocol_adapter",
+            "tool_protocol_adapter.bridge",
+            "toolspec_bridge",
+            "toolspec_bridge.adapter",
+            "toolspec_head",
+            "toolspec_head.param_schema_adapter",
+            "toolspec_head.postcondition_adapter",
+            "toolspec_head.precondition_adapter",
+            "toolspec_head.receipt_formatter",
+            "toolspec_head.trigger_gate",
+            "toolspec_head.validator_gate",
+            "validator_feedback_loop.rollback_adapter"
+          ],
+          "name": "group_2_custom_runtime_harness_bridge",
+          "rank": 8,
+          "rationale": "Runtime / harness obedience should attach to explicit interface bridges before any backbone fallback.",
+          "risk": "low",
+          "surface": "custom_runtime_harness_bridge"
+        },
+        {
+          "alpha": 16,
+          "dropout": 0.0,
+          "exclude_patterns": [
+            "(^|\\.)lm_head($|\\.)",
+            "(^|\\.)embed_tokens($|\\.)",
+            "(^|\\.)norm($|\\.)",
+            "attnres|attention_residual",
+            "dualpath",
+            "mhc|hyper[-_]?connection",
+            "rope|rotary",
+            "kvcache|kv_cache",
+            "(^|\\.)memory(_store|_bank)?($|\\.)"
+          ],
+          "include_patterns": [
+            "(^|\\.)(validator_feedback|sandbox_result_validator|sandbox_rollback_bridge|failure_signal|rollback)(_adapter|_bridge|_head)?($|\\.)"
+          ],
+          "intents": [
+            "validator_alignment"
+          ],
+          "layer_window": {
+            "mode": "all",
+            "value": 0.0
+          },
+          "matched_module_names": [
+            "failure_signal_bridge",
+            "failure_signal_bridge.rollback_bridge",
+            "rollback_adapter",
+            "rollback_adapter.head",
+            "sandbox_rollback_bridge",
+            "sandbox_rollback_bridge.adapter",
+            "validator_feedback_bridge",
+            "validator_feedback_bridge.adapter",
+            "validator_feedback_loop.rollback_adapter"
+          ],
+          "name": "group_3_custom_validator_bridge",
+          "rank": 8,
+          "rationale": "Validation and rollback fidelity should prefer explicit validator / rollback bridges.",
+          "risk": "low",
+          "surface": "custom_validator_bridge"
+        },
+        {
+          "alpha": 8,
+          "dropout": 0.0,
+          "exclude_patterns": [
+            "(^|\\.)lm_head($|\\.)",
+            "(^|\\.)embed_tokens($|\\.)",
+            "(^|\\.)norm($|\\.)",
+            "attnres|attention_residual",
+            "dualpath",
+            "mhc|hyper[-_]?connection",
+            "rope|rotary",
+            "kvcache|kv_cache",
+            "(^|\\.)memory(_store|_bank)?($|\\.)"
+          ],
+          "include_patterns": [
+            "(^|\\.)(memory_boundary_guard|episodic_memory|session_compactor)(_adapter|_bridge|_head)?($|\\.)"
+          ],
+          "intents": [
+            "session_continuity"
+          ],
+          "layer_window": {
+            "mode": "all",
+            "value": 0.0
+          },
+          "matched_module_names": [
+            "episodic_memory",
+            "episodic_memory.adapter",
+            "memory_boundary_guard",
+            "memory_boundary_guard.adapter",
+            "memory_boundary_guard.rollback_filter",
+            "session_compactor",
+            "session_compactor.adapter"
+          ],
+          "name": "group_4_custom_memory_boundary_bridge",
+          "rank": 4,
+          "rationale": "Session continuity should bind to boundary-aware memory packet surfaces rather than broad backbone tuning.",
+          "risk": "low",
+          "surface": "custom_memory_boundary_bridge"
+        }
+      ],
+      "upstream_profile_hint": null,
+      "version": "veriloop.peft_target_selector.v2",
+      "warnings": []
+    },
+    "training_mode": "mounted_head",
+    "version": "veriloop.uncertainty_adapter_trainer.v5.qwen36",
+    "warnings": [
+      "Harness Engineering is primary; PEFT is limited to obedience-facing, interface-facing support surfaces.",
+      "Backbone bridge tuning disabled explicitly; selector stays on custom surfaces or no-op."
+    ]
+  },
+  "status": "trained",
+  "train_metrics": {
+    "adapter_exported": true,
+    "auto_lora_from_ia3": false,
+    "best_epoch": 2,
+    "best_quality_score": 0.8625935807221907,
+    "epochs_completed": 4,
+    "loss": 0.009006613283418119,
+    "micro_batches": 154,
+    "micro_batches_total": 616,
+    "optimizer_steps": 10,
+    "optimizer_steps_total": 40,
+    "peft_method": "lora_narrow",
+    "used_peft": true
+  },
+  "version": "veriloop.uncertainty_adapter_trainer.v5.qwen36",
+  "warnings": [
+    "Harness Engineering is primary; PEFT is limited to obedience-facing, interface-facing support surfaces.",
+    "Backbone bridge tuning disabled explicitly; selector stays on custom surfaces or no-op.",
+    "Synthetic holdout split used for quality-oriented train/eval separation."
+  ]
+}

uncertainty_eval.jsonl ADDED Viewed

	@@ -0,0 +1,55 @@

+{"mode": "low_uncertainty", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: multiple corroborating sources agree and carry stable provenance.\nValidation state: latest validator receipt succeeded without rollback.\nWorktree state: clean snapshot with no conflicting edit.\nPatch state: none pending.\nToolspec state: bounded call contract is fully aligned.", "target": {"u_answer": 0.08, "u_evidence": 0.1, "u_exec": 0.06, "u_spec": 0.05, "u_risk": 0.08}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "low_uncertainty", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: multiple corroborating sources agree and carry stable provenance.\nValidation state: latest validator receipt succeeded without rollback.\nWorktree state: clean snapshot with no conflicting edit.\nPatch state: none pending.\nToolspec state: bounded call contract is fully aligned.", "target": {"u_answer": 0.08, "u_evidence": 0.1, "u_exec": 0.06, "u_spec": 0.05, "u_risk": 0.08}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "low_uncertainty", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: multiple corroborating sources agree and carry stable provenance.\nValidation state: latest validator receipt succeeded without rollback.\nWorktree state: clean snapshot with no conflicting edit.\nPatch state: none pending.\nToolspec state: bounded call contract is fully aligned.", "target": {"u_answer": 0.08, "u_evidence": 0.1, "u_exec": 0.06, "u_spec": 0.05, "u_risk": 0.08}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "low_uncertainty", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: multiple corroborating sources agree and carry stable provenance.\nValidation state: latest validator receipt succeeded without rollback.\nWorktree state: clean snapshot with no conflicting edit.\nPatch state: none pending.\nToolspec state: bounded call contract is fully aligned.", "target": {"u_answer": 0.08, "u_evidence": 0.1, "u_exec": 0.06, "u_spec": 0.05, "u_risk": 0.08}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "low_uncertainty", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: multiple corroborating sources agree and carry stable provenance.\nValidation state: latest validator receipt succeeded without rollback.\nWorktree state: clean snapshot with no conflicting edit.\nPatch state: none pending.\nToolspec state: bounded call contract is fully aligned.", "target": {"u_answer": 0.08, "u_evidence": 0.1, "u_exec": 0.06, "u_spec": 0.05, "u_risk": 0.08}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "evidence_gap", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: retrieval produced sparse evidence and citation confidence remains low.\nValidation state: no explicit failure yet, but claim support is visibly underdetermined.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: currently legal.", "target": {"u_answer": 0.42, "u_evidence": 0.86, "u_exec": 0.38, "u_spec": 0.18, "u_risk": 0.58}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "evidence_gap", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: retrieval produced sparse evidence and citation confidence remains low.\nValidation state: no explicit failure yet, but claim support is visibly underdetermined.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: currently legal.", "target": {"u_answer": 0.42, "u_evidence": 0.86, "u_exec": 0.38, "u_spec": 0.18, "u_risk": 0.58}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "evidence_gap", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: retrieval produced sparse evidence and citation confidence remains low.\nValidation state: no explicit failure yet, but claim support is visibly underdetermined.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: currently legal.", "target": {"u_answer": 0.42, "u_evidence": 0.86, "u_exec": 0.38, "u_spec": 0.18, "u_risk": 0.58}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "evidence_gap", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: retrieval produced sparse evidence and citation confidence remains low.\nValidation state: no explicit failure yet, but claim support is visibly underdetermined.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: currently legal.", "target": {"u_answer": 0.42, "u_evidence": 0.86, "u_exec": 0.38, "u_spec": 0.18, "u_risk": 0.58}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "evidence_gap", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: retrieval produced sparse evidence and citation confidence remains low.\nValidation state: no explicit failure yet, but claim support is visibly underdetermined.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: currently legal.", "target": {"u_answer": 0.42, "u_evidence": 0.86, "u_exec": 0.38, "u_spec": 0.18, "u_risk": 0.58}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "exec_required", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: reasoning narrows the options but cannot finalize action legality.\nValidation state: no execution receipt exists yet.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: valid once execution is routed.", "target": {"u_answer": 0.48, "u_evidence": 0.52, "u_exec": 0.94, "u_spec": 0.16, "u_risk": 0.55}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "exec_required", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: reasoning narrows the options but cannot finalize action legality.\nValidation state: no execution receipt exists yet.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: valid once execution is routed.", "target": {"u_answer": 0.48, "u_evidence": 0.52, "u_exec": 0.94, "u_spec": 0.16, "u_risk": 0.55}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "exec_required", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: reasoning narrows the options but cannot finalize action legality.\nValidation state: no execution receipt exists yet.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: valid once execution is routed.", "target": {"u_answer": 0.48, "u_evidence": 0.52, "u_exec": 0.94, "u_spec": 0.16, "u_risk": 0.55}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "exec_required", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: reasoning narrows the options but cannot finalize action legality.\nValidation state: no execution receipt exists yet.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: valid once execution is routed.", "target": {"u_answer": 0.48, "u_evidence": 0.52, "u_exec": 0.94, "u_spec": 0.16, "u_risk": 0.55}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "exec_required", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: reasoning narrows the options but cannot finalize action legality.\nValidation state: no execution receipt exists yet.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: valid once execution is routed.", "target": {"u_answer": 0.48, "u_evidence": 0.52, "u_exec": 0.94, "u_spec": 0.16, "u_risk": 0.55}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "spec_mismatch", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: answer content is mostly grounded.\nValidation state: validator rejects the call because required fields do not match the live toolspec.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: contract mismatch remains unresolved.", "target": {"u_answer": 0.34, "u_evidence": 0.28, "u_exec": 0.74, "u_spec": 0.96, "u_risk": 0.62}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "spec_mismatch", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: answer content is mostly grounded.\nValidation state: validator rejects the call because required fields do not match the live toolspec.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: contract mismatch remains unresolved.", "target": {"u_answer": 0.34, "u_evidence": 0.28, "u_exec": 0.74, "u_spec": 0.96, "u_risk": 0.62}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "spec_mismatch", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: answer content is mostly grounded.\nValidation state: validator rejects the call because required fields do not match the live toolspec.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: contract mismatch remains unresolved.", "target": {"u_answer": 0.34, "u_evidence": 0.28, "u_exec": 0.74, "u_spec": 0.96, "u_risk": 0.62}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "spec_mismatch", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: answer content is mostly grounded.\nValidation state: validator rejects the call because required fields do not match the live toolspec.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: contract mismatch remains unresolved.", "target": {"u_answer": 0.34, "u_evidence": 0.28, "u_exec": 0.74, "u_spec": 0.96, "u_risk": 0.62}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "spec_mismatch", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: answer content is mostly grounded.\nValidation state: validator rejects the call because required fields do not match the live toolspec.\nWorktree state: unchanged.\nPatch state: none.\nToolspec state: contract mismatch remains unresolved.", "target": {"u_answer": 0.34, "u_evidence": 0.28, "u_exec": 0.74, "u_spec": 0.96, "u_risk": 0.62}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "high_risk", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: only moderate support is present.\nValidation state: no execution or validator adjudication has reduced the risk yet.\nWorktree state: unchanged.\nPatch state: none.\nRisk note: unsupported output would cause material downstream harm.", "target": {"u_answer": 0.6, "u_evidence": 0.7, "u_exec": 0.48, "u_spec": 0.3, "u_risk": 0.99}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "high_risk", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: only moderate support is present.\nValidation state: no execution or validator adjudication has reduced the risk yet.\nWorktree state: unchanged.\nPatch state: none.\nRisk note: unsupported output would cause material downstream harm.", "target": {"u_answer": 0.6, "u_evidence": 0.7, "u_exec": 0.48, "u_spec": 0.3, "u_risk": 0.99}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "high_risk", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: only moderate support is present.\nValidation state: no execution or validator adjudication has reduced the risk yet.\nWorktree state: unchanged.\nPatch state: none.\nRisk note: unsupported output would cause material downstream harm.", "target": {"u_answer": 0.6, "u_evidence": 0.7, "u_exec": 0.48, "u_spec": 0.3, "u_risk": 0.99}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "high_risk", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: only moderate support is present.\nValidation state: no execution or validator adjudication has reduced the risk yet.\nWorktree state: unchanged.\nPatch state: none.\nRisk note: unsupported output would cause material downstream harm.", "target": {"u_answer": 0.6, "u_evidence": 0.7, "u_exec": 0.48, "u_spec": 0.3, "u_risk": 0.99}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "high_risk", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: only moderate support is present.\nValidation state: no execution or validator adjudication has reduced the risk yet.\nWorktree state: unchanged.\nPatch state: none.\nRisk note: unsupported output would cause material downstream harm.", "target": {"u_answer": 0.6, "u_evidence": 0.7, "u_exec": 0.48, "u_spec": 0.3, "u_risk": 0.99}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "conflicting_evidence", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: one source supports the claim while another equally credible source negates it.\nValidation state: validator has not broken the tie.\nWorktree state: unchanged.\nPatch state: none.\nConflict note: evidence graph still contains unresolved contradiction edges.", "target": {"u_answer": 0.78, "u_evidence": 0.82, "u_exec": 0.46, "u_spec": 0.22, "u_risk": 0.74}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "conflicting_evidence", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: one source supports the claim while another equally credible source negates it.\nValidation state: validator has not broken the tie.\nWorktree state: unchanged.\nPatch state: none.\nConflict note: evidence graph still contains unresolved contradiction edges.", "target": {"u_answer": 0.78, "u_evidence": 0.82, "u_exec": 0.46, "u_spec": 0.22, "u_risk": 0.74}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "conflicting_evidence", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: one source supports the claim while another equally credible source negates it.\nValidation state: validator has not broken the tie.\nWorktree state: unchanged.\nPatch state: none.\nConflict note: evidence graph still contains unresolved contradiction edges.", "target": {"u_answer": 0.78, "u_evidence": 0.82, "u_exec": 0.46, "u_spec": 0.22, "u_risk": 0.74}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "conflicting_evidence", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: one source supports the claim while another equally credible source negates it.\nValidation state: validator has not broken the tie.\nWorktree state: unchanged.\nPatch state: none.\nConflict note: evidence graph still contains unresolved contradiction edges.", "target": {"u_answer": 0.78, "u_evidence": 0.82, "u_exec": 0.46, "u_spec": 0.22, "u_risk": 0.74}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "conflicting_evidence", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: one source supports the claim while another equally credible source negates it.\nValidation state: validator has not broken the tie.\nWorktree state: unchanged.\nPatch state: none.\nConflict note: evidence graph still contains unresolved contradiction edges.", "target": {"u_answer": 0.78, "u_evidence": 0.82, "u_exec": 0.46, "u_spec": 0.22, "u_risk": 0.74}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "validator_negation", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: previous support is now contested by a failing validator receipt.\nValidation state: latest validator explicitly rejects the active state.\nWorktree state: patch remains mounted.\nPatch state: rollback or repair decision is still pending.\nFailure route: rollback bridge is armed.", "target": {"u_answer": 0.86, "u_evidence": 0.64, "u_exec": 0.72, "u_spec": 0.2, "u_risk": 0.8}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "validator_negation", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: previous support is now contested by a failing validator receipt.\nValidation state: latest validator explicitly rejects the active state.\nWorktree state: patch remains mounted.\nPatch state: rollback or repair decision is still pending.\nFailure route: rollback bridge is armed.", "target": {"u_answer": 0.86, "u_evidence": 0.64, "u_exec": 0.72, "u_spec": 0.2, "u_risk": 0.8}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "validator_negation", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: previous support is now contested by a failing validator receipt.\nValidation state: latest validator explicitly rejects the active state.\nWorktree state: patch remains mounted.\nPatch state: rollback or repair decision is still pending.\nFailure route: rollback bridge is armed.", "target": {"u_answer": 0.86, "u_evidence": 0.64, "u_exec": 0.72, "u_spec": 0.2, "u_risk": 0.8}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "validator_negation", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: previous support is now contested by a failing validator receipt.\nValidation state: latest validator explicitly rejects the active state.\nWorktree state: patch remains mounted.\nPatch state: rollback or repair decision is still pending.\nFailure route: rollback bridge is armed.", "target": {"u_answer": 0.86, "u_evidence": 0.64, "u_exec": 0.72, "u_spec": 0.2, "u_risk": 0.8}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "validator_negation", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: previous support is now contested by a failing validator receipt.\nValidation state: latest validator explicitly rejects the active state.\nWorktree state: patch remains mounted.\nPatch state: rollback or repair decision is still pending.\nFailure route: rollback bridge is armed.", "target": {"u_answer": 0.86, "u_evidence": 0.64, "u_exec": 0.72, "u_spec": 0.2, "u_risk": 0.8}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "worktree_conflict", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: moderate support.\nValidation state: validator cannot finalize because the workspace is not in a coherent state.\nWorktree state: overlapping edits violate expected snapshot continuity.\nPatch state: pending.\nPermission note: bounded write discipline is still active.", "target": {"u_answer": 0.44, "u_evidence": 0.34, "u_exec": 0.7, "u_spec": 0.78, "u_risk": 0.68}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "worktree_conflict", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: moderate support.\nValidation state: validator cannot finalize because the workspace is not in a coherent state.\nWorktree state: overlapping edits violate expected snapshot continuity.\nPatch state: pending.\nPermission note: bounded write discipline is still active.", "target": {"u_answer": 0.44, "u_evidence": 0.34, "u_exec": 0.7, "u_spec": 0.78, "u_risk": 0.68}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "worktree_conflict", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: moderate support.\nValidation state: validator cannot finalize because the workspace is not in a coherent state.\nWorktree state: overlapping edits violate expected snapshot continuity.\nPatch state: pending.\nPermission note: bounded write discipline is still active.", "target": {"u_answer": 0.44, "u_evidence": 0.34, "u_exec": 0.7, "u_spec": 0.78, "u_risk": 0.68}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "worktree_conflict", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: moderate support.\nValidation state: validator cannot finalize because the workspace is not in a coherent state.\nWorktree state: overlapping edits violate expected snapshot continuity.\nPatch state: pending.\nPermission note: bounded write discipline is still active.", "target": {"u_answer": 0.44, "u_evidence": 0.34, "u_exec": 0.7, "u_spec": 0.78, "u_risk": 0.68}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "worktree_conflict", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: moderate support.\nValidation state: validator cannot finalize because the workspace is not in a coherent state.\nWorktree state: overlapping edits violate expected snapshot continuity.\nPatch state: pending.\nPermission note: bounded write discipline is still active.", "target": {"u_answer": 0.44, "u_evidence": 0.34, "u_exec": 0.7, "u_spec": 0.78, "u_risk": 0.68}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "patch_pending", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: adequate for a repair attempt but not yet for final acceptance.\nValidation state: validator requested one more bounded patch cycle.\nWorktree state: stable.\nPatch state: corrective artifact exists but remains unverified.\nReceipt state: final acceptance receipt not yet produced.", "target": {"u_answer": 0.38, "u_evidence": 0.36, "u_exec": 0.66, "u_spec": 0.26, "u_risk": 0.5}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "patch_pending", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: adequate for a repair attempt but not yet for final acceptance.\nValidation state: validator requested one more bounded patch cycle.\nWorktree state: stable.\nPatch state: corrective artifact exists but remains unverified.\nReceipt state: final acceptance receipt not yet produced.", "target": {"u_answer": 0.38, "u_evidence": 0.36, "u_exec": 0.66, "u_spec": 0.26, "u_risk": 0.5}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "patch_pending", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: adequate for a repair attempt but not yet for final acceptance.\nValidation state: validator requested one more bounded patch cycle.\nWorktree state: stable.\nPatch state: corrective artifact exists but remains unverified.\nReceipt state: final acceptance receipt not yet produced.", "target": {"u_answer": 0.38, "u_evidence": 0.36, "u_exec": 0.66, "u_spec": 0.26, "u_risk": 0.5}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "patch_pending", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: adequate for a repair attempt but not yet for final acceptance.\nValidation state: validator requested one more bounded patch cycle.\nWorktree state: stable.\nPatch state: corrective artifact exists but remains unverified.\nReceipt state: final acceptance receipt not yet produced.", "target": {"u_answer": 0.38, "u_evidence": 0.36, "u_exec": 0.66, "u_spec": 0.26, "u_risk": 0.5}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "patch_pending", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: adequate for a repair attempt but not yet for final acceptance.\nValidation state: validator requested one more bounded patch cycle.\nWorktree state: stable.\nPatch state: corrective artifact exists but remains unverified.\nReceipt state: final acceptance receipt not yet produced.", "target": {"u_answer": 0.38, "u_evidence": 0.36, "u_exec": 0.66, "u_spec": 0.26, "u_risk": 0.5}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "self_check_failure", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: prior reasoning looked sound.\nValidation state: the self-check loop surfaced a failing test after generation.\nWorktree state: stable.\nPatch state: last edit remains mounted.\nInterpretation: submit would be premature without another bounded repair loop.", "target": {"u_answer": 0.55, "u_evidence": 0.3, "u_exec": 0.88, "u_spec": 0.22, "u_risk": 0.63}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "self_check_failure", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: prior reasoning looked sound.\nValidation state: the self-check loop surfaced a failing test after generation.\nWorktree state: stable.\nPatch state: last edit remains mounted.\nInterpretation: submit would be premature without another bounded repair loop.", "target": {"u_answer": 0.55, "u_evidence": 0.3, "u_exec": 0.88, "u_spec": 0.22, "u_risk": 0.63}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "self_check_failure", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: decide whether a post-generation self-check loop is still required.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: prior reasoning looked sound.\nValidation state: the self-check loop surfaced a failing test after generation.\nWorktree state: stable.\nPatch state: last edit remains mounted.\nInterpretation: submit would be premature without another bounded repair loop.", "target": {"u_answer": 0.55, "u_evidence": 0.3, "u_exec": 0.88, "u_spec": 0.22, "u_risk": 0.63}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "self_check_failure", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: estimate runtime risk before trusting a partially-validated repository change.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: prior reasoning looked sound.\nValidation state: the self-check loop surfaced a failing test after generation.\nWorktree state: stable.\nPatch state: last edit remains mounted.\nInterpretation: submit would be premature without another bounded repair loop.", "target": {"u_answer": 0.55, "u_evidence": 0.3, "u_exec": 0.88, "u_spec": 0.22, "u_risk": 0.63}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "self_check_failure", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: prior reasoning looked sound.\nValidation state: the self-check loop surfaced a failing test after generation.\nWorktree state: stable.\nPatch state: last edit remains mounted.\nInterpretation: submit would be premature without another bounded repair loop.", "target": {"u_answer": 0.55, "u_evidence": 0.3, "u_exec": 0.88, "u_spec": 0.22, "u_risk": 0.63}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}
+{"mode": "reverse_engineering_ambiguity", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #1.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: repository topology is only partially reconstructed.\nValidation state: no sandbox failure exists yet, but architecture intent is not resolved.\nWorktree state: unchanged.\nPatch state: none.\nInterpretation: reverse-engineering readiness is required before legal mutation.", "target": {"u_answer": 0.52, "u_evidence": 0.76, "u_exec": 0.68, "u_spec": 0.72, "u_risk": 0.44}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 0}}
+{"mode": "reverse_engineering_ambiguity", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #2.\nTask brief: decide whether another sandbox run is needed before accepting a patch.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: repository topology is only partially reconstructed.\nValidation state: no sandbox failure exists yet, but architecture intent is not resolved.\nWorktree state: unchanged.\nPatch state: none.\nInterpretation: reverse-engineering readiness is required before legal mutation.", "target": {"u_answer": 0.52, "u_evidence": 0.76, "u_exec": 0.68, "u_spec": 0.72, "u_risk": 0.44}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 1}}
+{"mode": "reverse_engineering_ambiguity", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #3.\nTask brief: infer repository intent when the codebase must be reverse engineered before editing.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: repository topology is only partially reconstructed.\nValidation state: no sandbox failure exists yet, but architecture intent is not resolved.\nWorktree state: unchanged.\nPatch state: none.\nInterpretation: reverse-engineering readiness is required before legal mutation.", "target": {"u_answer": 0.52, "u_evidence": 0.76, "u_exec": 0.68, "u_spec": 0.72, "u_risk": 0.44}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 2}}
+{"mode": "reverse_engineering_ambiguity", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #4.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: repository topology is only partially reconstructed.\nValidation state: no sandbox failure exists yet, but architecture intent is not resolved.\nWorktree state: unchanged.\nPatch state: none.\nInterpretation: reverse-engineering readiness is required before legal mutation.", "target": {"u_answer": 0.52, "u_evidence": 0.76, "u_exec": 0.68, "u_spec": 0.72, "u_risk": 0.44}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 3}}
+{"mode": "reverse_engineering_ambiguity", "product_line": "veriloop_coder", "prompt": "Evaluation scenario #5.\nTask brief: judge whether a code patch is ready to submit after validator feedback.\nProduct line: veriloop_coder.\nRuntime mode: verify-before-submit.\nHarness state: validator loop active; permission context active; worktree state tracked; runtime protocol active.\nReturn uncertainty only as a structured five-dimensional estimate.\nEvidence bundle: repository topology is only partially reconstructed.\nValidation state: no sandbox failure exists yet, but architecture intent is not resolved.\nWorktree state: unchanged.\nPatch state: none.\nInterpretation: reverse-engineering readiness is required before legal mutation.", "target": {"u_answer": 0.52, "u_evidence": 0.76, "u_exec": 0.68, "u_spec": 0.72, "u_risk": 0.44}, "metadata": {"synthetic": true, "split": "eval", "sample_idx": 4}}

uncertainty_head.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:bdd4782fe5d2f26c184c8690a07527cffa99eebe763e415a4ef449ed8046c88a
+size 25220609

uncertainty_train.jsonl ADDED Viewed

The diff for this file is too large to render. See raw diff

uncertainty_training_manifest.json ADDED Viewed

	@@ -0,0 +1,105 @@

+{
+  "adapter_exported": true,
+  "auto_lora_from_ia3": false,
+  "best_epoch": 2,
+  "best_quality_score": 0.8625935807221907,
+  "dataset_summary": {
+    "eval_modes": {
+      "conflicting_evidence": 5,
+      "evidence_gap": 5,
+      "exec_required": 5,
+      "high_risk": 5,
+      "low_uncertainty": 5,
+      "patch_pending": 5,
+      "reverse_engineering_ambiguity": 5,
+      "self_check_failure": 5,
+      "spec_mismatch": 5,
+      "validator_negation": 5,
+      "worktree_conflict": 5
+    },
+    "eval_size": 55,
+    "train_modes": {
+      "conflicting_evidence": 14,
+      "evidence_gap": 14,
+      "exec_required": 14,
+      "high_risk": 14,
+      "low_uncertainty": 14,
+      "patch_pending": 14,
+      "reverse_engineering_ambiguity": 14,
+      "self_check_failure": 14,
+      "spec_mismatch": 14,
+      "validator_negation": 14,
+      "worktree_conflict": 14
+    },
+    "train_size": 154
+  },
+  "epochs_completed": 4,
+  "eval_metrics": {
+    "adapter_exported": true,
+    "auto_lora_from_ia3": false,
+    "best_epoch": 2,
+    "best_quality_score": 0.8625935807221907,
+    "count": 55,
+    "mae": {
+      "u_answer": 0.15174226462841034,
+      "u_evidence": 0.19610758125782013,
+      "u_exec": 0.18561214208602905,
+      "u_risk": 0.1553734689950943,
+      "u_spec": 0.21633382141590118
+    },
+    "mean_mae": 0.18103384971618652,
+    "mean_rmse": 0.24169571697711945,
+    "moderate_accuracy": 0.6727272727272727,
+    "peft_method": "lora_narrow",
+    "quality_score": 0.8625935807221907,
+    "rmse": {
+      "u_answer": 0.18724055588245392,
+      "u_evidence": 0.22527915239334106,
+      "u_exec": 0.25238174200057983,
+      "u_risk": 0.20667441189289093,
+      "u_spec": 0.3369026482105255
+    },
+    "tight_accuracy": 0.4,
+    "used_peft": true,
+    "weighted_mae": 0.18083095811830807,
+    "weighted_rmse": 0.24125460771003793
+  },
+  "load_meta": {
+    "chosen_class": "AutoModelForCausalLM",
+    "hidden_size": 2048,
+    "quantization_mode": "4bit"
+  },
+  "peft_method": "lora_narrow",
+  "requested_method": "lora_narrow",
+  "requested_target_modules": [
+    "uncertainty_head",
+    "uncertainty_head.calibration_mlp",
+    "uncertainty_head.proj"
+  ],
+  "resolved_target_modules": [
+    "surface_host.uncertainty_head.calibration_mlp",
+    "surface_host.uncertainty_head.proj"
+  ],
+  "selected_target_modules": [
+    "uncertainty_head",
+    "uncertainty_head.calibration_mlp",
+    "uncertainty_head.proj"
+  ],
+  "status": "trained",
+  "train_metrics": {
+    "adapter_exported": true,
+    "auto_lora_from_ia3": false,
+    "best_epoch": 2,
+    "best_quality_score": 0.8625935807221907,
+    "epochs_completed": 4,
+    "loss": 0.009006613283418119,
+    "micro_batches": 154,
+    "micro_batches_total": 616,
+    "optimizer_steps": 10,
+    "optimizer_steps_total": 40,
+    "peft_method": "lora_narrow",
+    "used_peft": true
+  },
+  "unresolved_target_modules": [],
+  "used_peft": true
+}