Add files using upload-large-folder tool

omni_diffusion/mechanism_cases/beach/confidence/omni_t2i_progressive_confidence.json

@@ -0,0 +1,37 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_last2_progressive_gate_20260730/progressive_confidence/checkpoint-500",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "Two little boys and a little girl playing on a sunny beach with clear blue skies, as they explore the shore and dig in the sand with their toy pail and shovel.",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "dprm_action_guidance_scale": 20.0,
+  "seed": 20270085,
+  "generation_seconds": 20.870100736618042,
+  "total_seconds": 38.40964365005493,
+  "num_image_tokens": 256,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0032.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0064.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0096.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0128.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0160.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0192.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0224.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0256.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/confidence/history_frames/step_0259.png"
+  ]
+}

omni_diffusion/mechanism_cases/beach/dprm/omni_t2i_progressive_confidence.json

@@ -0,0 +1,76 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_last2_progressive_gate_20260730/progressive_confidence/checkpoint-500",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "Two little boys and a little girl playing on a sunny beach with clear blue skies, as they explore the shore and dig in the sand with their toy pail and shovel.",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "dprm_action_guidance_scale": 20.0,
+  "seed": 20270085,
+  "generation_seconds": 24.3183913230896,
+  "total_seconds": 42.999722957611084,
+  "num_image_tokens": 256,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0032.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0064.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0096.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0128.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0160.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0192.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0224.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0256.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_intermediate_canvas_kids_20260812/dprm/history_frames/step_0259.png"
+  ],
+  "counterfactual_override": {
+    "applied": true,
+    "step": 96,
+    "requested_quantile": 0.15,
+    "candidate_count": 163,
+    "candidate_index": 137,
+    "sequence_position": 283,
+    "visual_index": 229,
+    "confidence": 0.020754337310791016,
+    "raw_order_score": -3.875,
+    "confidence_bin": 0,
+    "rank_quantile": 0.14814814925193787,
+    "rank_bin": 1,
+    "aux_bin": 13,
+    "phase": 2,
+    "provisional_token_id": 168076,
+    "default_candidate_indices": [
+      75
+    ],
+    "default_candidate_index": 75,
+    "default_sequence_position": 210,
+    "default_visual_index": 156,
+    "default_confidence": 0.03164658322930336,
+    "default_raw_order_score": -3.453125,
+    "default_rank_quantile": 1.0,
+    "confidence_gap_from_default": -0.010892245918512344,
+    "overridden_candidate_indices": [
+      137
+    ],
+    "row": 14,
+    "column": 5,
+    "row_normalized": 0.9333333333333333,
+    "column_normalized": 0.3333333333333333,
+    "center_distance": 6.96419413859206,
+    "local_neighbor_count": 8,
+    "local_revealed_neighbor_count": 0,
+    "local_revealed_fraction": 0.0,
+    "nearest_revealed_manhattan": 4
+  }
+}

omni_diffusion/mechanism_cases/boy_kittens/confidence/omni_t2i_progressive_confidence.json

@@ -0,0 +1,38 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_last2_progressive_gate_20260730/progressive_confidence/checkpoint-500",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "A boy sits on a stoop while playing a flute for an audience of kittens on the sidewalk.",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20270027,
+  "generation_seconds": 74.42293643951416,
+  "total_seconds": 92.43578934669495,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": false,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0032.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0064.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0096.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0128.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0192.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0224.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0256.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0259.png"
+  ]
+}

omni_diffusion/mechanism_cases/boy_kittens/confidence/run.log

@@ -0,0 +1,108 @@
+[2026-08-14 22:52:04,660] [INFO] [real_accelerator.py:222:get_accelerator] Setting ds_accelerator to cuda (auto detect)
+/home/nitanda_sub/mark/env/lib/python3.10/site-packages/pydub/utils.py:170: RuntimeWarning: Couldn't find ffmpeg or avconv - defaulting to ffmpeg, but may not work
+  warn("Couldn't find ffmpeg or avconv - defaulting to ffmpeg, but may not work", RuntimeWarning)
+WARNING:root:trust_remote_code: True
+Notice: ffmpeg is not installed. torchaudio is used to load audio
+If you want to use ffmpeg backend to load audio, please install it by:
+	sudo apt install ffmpeg # ubuntu
+	# brew install ffmpeg # mac
+Detect model requirements, begin to install it: /home/nitanda_sub/mark/.cache/huggingface/hub/models--FunAudioLLM--SenseVoiceSmall/snapshots/3847d57b6bdf2dd8875cb1508d2af43d80a16bf7/requirements.txt
+fail to install model requirements! 
+error error: externally-managed-environment
+
+× This environment is externally managed
+╰─> To install Python packages system-wide, try apt install
+    python3-xyz, where xyz is the package you are trying to
+    install.
+    
+    If you wish to install a non-Debian-packaged Python package,
+    create a virtual environment using python3 -m venv path/to/venv.
+    Then use path/to/venv/bin/python and path/to/venv/bin/pip. Make
+    sure you have python3-full installed.
+    
+    If you wish to install a non-Debian packaged Python application,
+    it may be easiest to use pipx install xyz, which will manage a
+    virtual environment for you. Make sure you have pipx installed.
+    
+    See /usr/share/doc/python3.12/README.venv for more information.
+
+note: If you believe this is a mistake, please contact your Python installation or OS distribution provider. You can override this, at the risk of breaking your Python installation or OS, by passing --break-system-packages.
+hint: See PEP 668 for the detailed specification.
+
+Loading remote code failed: model, No module named 'model'
+
+WARNING:root:trust_remote_code: True
+Detect model requirements, begin to install it: /home/nitanda_sub/mark/.cache/huggingface/hub/models--FunAudioLLM--SenseVoiceSmall/snapshots/3847d57b6bdf2dd8875cb1508d2af43d80a16bf7/requirements.txt
+fail to install model requirements! 
+error error: externally-managed-environment
+
+× This environment is externally managed
+╰─> To install Python packages system-wide, try apt install
+    python3-xyz, where xyz is the package you are trying to
+    install.
+    
+    If you wish to install a non-Debian-packaged Python package,
+    create a virtual environment using python3 -m venv path/to/venv.
+    Then use path/to/venv/bin/python and path/to/venv/bin/pip. Make
+    sure you have python3-full installed.
+    
+    If you wish to install a non-Debian packaged Python application,
+    it may be easiest to use pipx install xyz, which will manage a
+    virtual environment for you. Make sure you have pipx installed.
+    
+    See /usr/share/doc/python3.12/README.venv for more information.
+
+note: If you believe this is a mistake, please contact your Python installation or OS distribution provider. You can override this, at the risk of breaking your Python installation or OS, by passing --break-system-packages.
+hint: See PEP 668 for the detailed specification.
+
+Loading remote code failed: model, No module named 'model'
+
+INFO:omni_diffusion.tokenizer_magvitv2:self.rank=None
+INFO:omni_diffusion.tokenizer_magvitv2:self.device='cuda:0' Loading MAGVITv2
+INFO:omni_diffusion.tokenizer_magvitv2:self.device='cuda:0' Loading MAGVITv2 done
+target_ratios [(1, 1), (1, 2), (2, 1), (3, 1), (1, 3), (2, 2), (4, 1), (1, 4), (5, 1), (1, 5), (1, 6), (6, 1), (3, 2), (2, 3), (7, 1), (1, 7), (4, 2), (2, 4), (1, 8), (8, 1), (1, 9), (3, 3), (9, 1), (2, 5), (5, 2), (10, 1), (1, 10), (11, 1), (1, 11), (12, 1), (3, 4), (4, 3), (1, 12), (6, 2), (2, 6)]
+possible_resolutions [[512, 512], [512, 1024], [1024, 512], [1536, 512], [512, 1536], [1024, 1024], [2048, 512], [512, 2048], [2560, 512], [512, 2560], [512, 3072], [3072, 512], [1536, 1024], [1024, 1536], [3584, 512], [512, 3584], [2048, 1024], [1024, 2048], [512, 4096], [4096, 512], [512, 4608], [1536, 1536], [4608, 512], [1024, 2560], [2560, 1024], [5120, 512], [512, 5120], [5632, 512], [512, 5632], [6144, 512], [1536, 2048], [2048, 1536], [512, 6144], [3072, 1024], [1024, 3072]]
+Working with z of shape (1, 13, 16, 16) = 3328 dimensions.
+Look-up free quantizer with codebook size: 8192
+Working with z of shape (1, 13, 16, 16) = 3328 dimensions.
+Look-up free quantizer with codebook size: 8192
+denoise algorithm: entropy-penalty
+
  0%|          | 0/260 [00:00<?, ?it/s]
  0%|          | 1/260 [00:00<00:47,  5.44it/s]
  1%|          | 3/260 [00:00<00:27,  9.40it/s]
  2%|▏         | 5/260 [00:00<00:23, 10.65it/s]
  3%|▎         | 7/260 [00:00<00:24, 10.18it/s]
  3%|▎         | 9/260 [00:00<00:23, 10.72it/s]
  4%|▍         | 11/260 [00:01<00:22, 10.95it/s]
  5%|▌         | 13/260 [00:01<00:22, 10.92it/s]
  6%|▌         | 15/260 [00:01<00:24, 10.05it/s]
  7%|▋         | 17/260 [00:01<00:23, 10.25it/s]
  7%|▋         | 19/260 [00:01<00:25,  9.57it/s]
  8%|▊         | 20/260 [00:02<00:24,  9.64it/s]
  8%|▊         | 22/260 [00:02<00:24,  9.73it/s]
  9%|▉         | 23/260 [00:02<00:24,  9.52it/s]
  9%|▉         | 24/260 [00:02<00:26,  8.87it/s]
 10%|▉         | 25/260 [00:02<00:25,  9.05it/s]
 10%|█         | 26/260 [00:02<00:25,  9.18it/s]
 10%|█         | 27/260 [00:02<00:25,  9.19it/s]
 11%|█         | 28/260 [00:02<00:27,  8.33it/s]
 11%|█         | 29/260 [00:03<00:27,  8.51it/s]
 12%|█▏        | 30/260 [00:03<00:26,  8.67it/s]
 12%|█▏        | 31/260 [00:03<00:26,  8.71it/s]
 12%|█▏        | 32/260 [00:03<00:28,  7.88it/s]
 13%|█▎        | 33/260 [00:03<00:28,  8.01it/s]
 13%|█▎        | 34/260 [00:03<00:27,  8.21it/s]
 13%|█▎        | 35/260 [00:03<00:29,  7.54it/s]
 14%|█▍        | 36/260 [00:03<00:28,  7.82it/s]
 14%|█▍        | 37/260 [00:04<00:27,  7.98it/s]
 15%|█▍        | 38/260 [00:04<00:30,  7.34it/s]
 15%|█▌        | 39/260 [00:04<00:29,  7.60it/s]
 15%|█▌        | 40/260 [00:04<00:28,  7.67it/s]
 16%|█▌        | 41/260 [00:04<00:30,  7.09it/s]
 16%|█▌        | 42/260 [00:04<00:29,  7.36it/s]
 17%|█▋        | 43/260 [00:04<00:28,  7.53it/s]
 17%|█▋        | 44/260 [00:05<00:30,  6.97it/s]
 17%|█▋        | 45/260 [00:05<00:29,  7.22it/s]
 18%|█▊        | 46/260 [00:05<00:28,  7.39it/s]
 18%|█▊        | 47/260 [00:05<00:31,  6.79it/s]
 18%|█▊        | 48/260 [00:05<00:30,  7.04it/s]
 19%|█▉        | 49/260 [00:05<00:29,  7.19it/s]
 19%|█▉        | 50/260 [00:05<00:31,  6.68it/s]
 20%|█▉        | 51/260 [00:06<00:30,  6.90it/s]
 20%|██        | 52/260 [00:06<00:29,  7.05it/s]
 20%|██        | 53/260 [00:06<00:31,  6.49it/s]
 21%|██        | 54/260 [00:06<00:30,  6.70it/s]
 21%|██        | 55/260 [00:06<00:32,  6.29it/s]
 22%|██▏       | 56/260 [00:06<00:31,  6.53it/s]
 22%|██▏       | 57/260 [00:06<00:30,  6.67it/s]
 22%|██▏       | 58/260 [00:07<00:32,  6.24it/s]
 23%|██▎       | 59/260 [00:07<00:31,  6.43it/s]
 23%|██▎       | 60/260 [00:07<00:33,  6.02it/s]
 23%|██▎       | 61/260 [00:07<00:31,  6.23it/s]
 24%|██▍       | 62/260 [00:07<00:31,  6.35it/s]
 24%|██▍       | 63/260 [00:08<00:55,  3.54it/s]
 25%|██▍       | 64/260 [00:08<00:58,  3.33it/s]
 25%|██▌       | 65/260 [00:08<00:52,  3.71it/s]
 25%|██▌       | 66/260 [00:09<00:45,  4.24it/s]
 26%|██▌       | 67/260 [00:09<00:43,  4.44it/s]
 26%|██▌       | 68/260 [00:09<00:39,  4.90it/s]
 27%|██▋       | 69/260 [00:09<00:36,  5.20it/s]
 27%|██▋       | 70/260 [00:09<00:36,  5.14it/s]
 27%|██▋       | 71/260 [00:09<00:35,  5.38it/s]
 28%|██▊       | 72/260 [00:10<00:36,  5.21it/s]
 28%|██▊       | 73/260 [00:10<00:34,  5.44it/s]
 28%|██▊       | 74/260 [00:10<00:35,  5.22it/s]
 29%|██▉       | 75/260 [00:10<00:34,  5.42it/s]
 29%|██▉       | 76/260 [00:10<00:35,  5.16it/s]
 30%|██▉       | 77/260 [00:11<00:34,  5.35it/s]
 30%|███       | 78/260 [00:11<00:36,  5.04it/s]
 30%|███       | 79/260 [00:11<00:34,  5.24it/s]
 31%|███       | 80/260 [00:11<00:35,  5.02it/s]
 31%|███       | 81/260 [00:11<00:34,  5.26it/s]
 32%|███▏      | 82/260 [00:12<00:35,  5.04it/s]
 32%|███▏      | 83/260 [00:12<00:33,  5.21it/s]
 32%|███▏      | 84/260 [00:12<00:35,  4.95it/s]
 33%|███▎      | 85/260 [00:12<00:34,  5.12it/s]
 33%|███▎      | 86/260 [00:12<00:35,  4.92it/s]
 33%|███▎      | 87/260 [00:13<00:34,  5.07it/s]
 34%|███▍      | 88/260 [00:13<00:35,  4.87it/s]
 34%|███▍      | 89/260 [00:13<00:34,  5.02it/s]
 35%|███▍      | 90/260 [00:13<00:35,  4.78it/s]
 35%|███▌      | 91/260 [00:13<00:34,  4.93it/s]
 35%|███▌      | 92/260 [00:14<00:35,  4.72it/s]
 36%|███▌      | 93/260 [00:14<00:36,  4.56it/s]
 36%|███▌      | 94/260 [00:14<00:35,  4.74it/s]
 37%|███▋      | 95/260 [00:14<00:35,  4.59it/s]
 37%|███▋      | 96/260 [00:14<00:34,  4.72it/s]
 37%|███▋      | 97/260 [00:15<00:35,  4.56it/s]
 38%|███▊      | 98/260 [00:15<00:34,  4.71it/s]
 38%|███▊      | 99/260 [00:15<00:35,  4.54it/s]
 38%|███▊      | 100/260 [00:15<00:34,  4.65it/s]
 39%|███▉      | 101/260 [00:16<00:35,  4.48it/s]
 39%|███▉      | 102/260 [00:16<00:36,  4.39it/s]
 40%|███▉      | 103/260 [00:16<00:34,  4.52it/s]
 40%|████      | 104/260 [00:16<00:35,  4.36it/s]
 40%|████      | 105/260 [00:16<00:34,  4.51it/s]
 41%|████      | 106/260 [00:17<00:35,  4.36it/s]
 41%|████      | 107/260 [00:17<00:34,  4.47it/s]
 42%|████▏     | 108/260 [00:17<00:35,  4.32it/s]
 42%|████▏     | 109/260 [00:17<00:35,  4.21it/s]
 42%|████▏     | 110/260 [00:18<00:34,  4.36it/s]
 43%|████▎     | 111/260 [00:18<00:35,  4.20it/s]
 43%|████▎     | 112/260 [00:18<00:34,  4.33it/s]
 43%|████▎     | 113/260 [00:18<00:35,  4.20it/s]
 44%|████▍     | 114/260 [00:19<00:35,  4.07it/s]
 44%|████▍     | 115/260 [00:19<00:34,  4.22it/s]
 45%|████▍     | 116/260 [00:19<00:35,  4.10it/s]
 45%|████▌     | 117/260 [00:19<00:33,  4.22it/s]
 45%|████▌     | 118/260 [00:20<00:34,  4.08it/s]
 46%|████▌     | 119/260 [00:20<00:35,  3.99it/s]
 46%|████▌     | 120/260 [00:20<00:34,  4.10it/s]
 47%|████▋     | 121/260 [00:20<00:34,  4.00it/s]
 47%|████▋     | 122/260 [00:21<00:34,  4.02it/s]
 47%|████▋     | 123/260 [00:21<00:34,  4.00it/s]
 48%|████▊     | 124/260 [00:21<00:35,  3.87it/s]
 48%|████▊     | 125/260 [00:21<00:35,  3.81it/s]
 48%|████▊     | 126/260 [00:22<00:34,  3.92it/s]
 49%|████▉     | 127/260 [00:22<00:34,  3.85it/s]
 49%|████▉     | 128/260 [00:22<00:34,  3.78it/s]
 50%|████▉     | 129/260 [00:22<00:33,  3.88it/s]
 50%|█████     | 130/260 [00:23<00:34,  3.78it/s]
 50%|█████     | 131/260 [00:23<00:33,  3.90it/s]
 51%|█████     | 132/260 [00:23<00:33,  3.78it/s]
 51%|█████     | 133/260 [00:24<00:34,  3.72it/s]
 52%|█████▏    | 134/260 [00:24<00:35,  3.57it/s]
 52%|█████▏    | 135/260 [00:24<00:33,  3.72it/s]
 52%|█████▏    | 136/260 [00:24<00:34,  3.65it/s]
 53%|█████▎    | 137/260 [00:25<00:34,  3.59it/s]
 53%|█████▎    | 138/260 [00:25<00:32,  3.72it/s]
 53%|█████▎    | 139/260 [00:25<00:33,  3.62it/s]
 54%|█████▍    | 140/260 [00:25<00:33,  3.55it/s]
 54%|█████▍    | 141/260 [00:26<00:32,  3.61it/s]
 55%|█████▍    | 142/260 [00:26<00:32,  3.60it/s]
 55%|█████▌    | 143/260 [00:26<00:33,  3.52it/s]
 55%|█████▌    | 144/260 [00:27<00:33,  3.47it/s]
 56%|█████▌    | 145/260 [00:27<00:31,  3.59it/s]
 56%|█████▌    | 146/260 [00:27<00:32,  3.51it/s]
 57%|█████▋    | 147/260 [00:27<00:32,  3.44it/s]
 57%|█████▋    | 148/260 [00:28<00:31,  3.54it/s]
 57%|█████▋    | 149/260 [00:28<00:31,  3.48it/s]
 58%|█████▊    | 150/260 [00:28<00:32,  3.41it/s]
 58%|█████▊    | 151/260 [00:29<00:32,  3.36it/s]
 58%|█████▊    | 152/260 [00:29<00:31,  3.46it/s]
 59%|█████▉    | 153/260 [00:29<00:31,  3.39it/s]
 59%|█████▉    | 154/260 [00:30<00:31,  3.34it/s]
 60%|█████▉    | 155/260 [00:30<00:31,  3.37it/s]
 60%|██████    | 156/260 [00:30<00:30,  3.40it/s]
 60%|██████    | 157/260 [00:30<00:30,  3.35it/s]
 61%|██████    | 158/260 [00:31<00:30,  3.29it/s]
 61%|██████    | 159/260 [00:31<00:29,  3.40it/s]
 62%|██████▏   | 160/260 [00:31<00:30,  3.32it/s]
 62%|██████▏   | 161/260 [00:32<00:29,  3.30it/s]
 62%|██████▏   | 162/260 [00:32<00:29,  3.27it/s]
 63%|██████▎   | 163/260 [00:32<00:28,  3.38it/s]
 63%|██████▎   | 164/260 [00:33<00:28,  3.31it/s]
 63%|██████▎   | 165/260 [00:33<00:29,  3.26it/s]
 64%|██████▍   | 166/260 [00:33<00:28,  3.35it/s]
 64%|██████▍   | 167/260 [00:33<00:28,  3.28it/s]
 65%|██████▍   | 168/260 [00:34<00:28,  3.23it/s]
 65%|██████▌   | 169/260 [00:34<00:28,  3.21it/s]
 65%|██████▌   | 170/260 [00:34<00:27,  3.30it/s]
 66%|██████▌   | 171/260 [00:35<00:27,  3.24it/s]
 66%|██████▌   | 172/260 [00:35<00:27,  3.20it/s]
 67%|██████▋   | 173/260 [00:35<00:27,  3.16it/s]
 67%|██████▋   | 174/260 [00:36<00:26,  3.24it/s]
 67%|██████▋   | 175/260 [00:36<00:26,  3.18it/s]
 68%|██████▊   | 176/260 [00:36<00:26,  3.13it/s]
 68%|██████▊   | 177/260 [00:37<00:26,  3.13it/s]
 68%|██████▊   | 178/260 [00:37<00:25,  3.19it/s]
 69%|██████▉   | 179/260 [00:37<00:26,  3.11it/s]
 69%|██████▉   | 180/260 [00:38<00:26,  3.07it/s]
 70%|██████▉   | 181/260 [00:38<00:25,  3.07it/s]
 70%|███████   | 182/260 [00:38<00:24,  3.13it/s]
 70%|███████   | 183/260 [00:39<00:25,  3.08it/s]
 71%|███████   | 184/260 [00:39<00:25,  3.04it/s]
 71%|███████   | 185/260 [00:39<00:24,  3.00it/s]
 72%|███████▏  | 186/260 [00:40<00:23,  3.08it/s]
 72%|███████▏  | 187/260 [00:40<00:24,  3.04it/s]
 72%|███████▏  | 188/260 [00:40<00:23,  3.01it/s]
 73%|███████▎  | 189/260 [00:41<00:23,  2.97it/s]
 73%|███████▎  | 190/260 [00:41<00:22,  3.05it/s]
 73%|███████▎  | 191/260 [00:41<00:23,  2.99it/s]
 74%|███████▍  | 192/260 [00:42<00:23,  2.95it/s]
 74%|███████▍  | 193/260 [00:42<00:22,  2.91it/s]
 75%|███████▍  | 194/260 [00:42<00:22,  2.99it/s]
 75%|███████▌  | 195/260 [00:43<00:22,  2.94it/s]
 75%|███████▌  | 196/260 [00:43<00:22,  2.90it/s]
 76%|███████▌  | 197/260 [00:43<00:21,  2.99it/s]
 76%|███████▌  | 198/260 [00:44<00:28,  2.17it/s]
 77%|███████▋  | 199/260 [00:45<00:31,  1.92it/s]
 77%|███████▋  | 200/260 [00:45<00:33,  1.82it/s]
 77%|███████▋  | 201/260 [00:46<00:32,  1.79it/s]
 78%|███████▊  | 202/260 [00:46<00:32,  1.77it/s]
 78%|███████▊  | 203/260 [00:47<00:32,  1.78it/s]
 78%|███████▊  | 204/260 [00:48<00:31,  1.79it/s]
 79%|███████▉  | 205/260 [00:48<00:29,  1.86it/s]
 79%|███████▉  | 206/260 [00:49<00:29,  1.84it/s]
 80%|███████▉  | 207/260 [00:49<00:27,  1.90it/s]
 80%|████████  | 208/260 [00:50<00:26,  1.93it/s]
 80%|████████  | 209/260 [00:50<00:26,  1.96it/s]
 81%|████████  | 210/260 [00:51<00:25,  1.98it/s]
 81%|████████  | 211/260 [00:51<00:24,  1.99it/s]
 82%|████████▏ | 212/260 [00:52<00:23,  2.00it/s]
 82%|████████▏ | 213/260 [00:52<00:22,  2.07it/s]
 82%|████████▏ | 214/260 [00:53<00:22,  2.06it/s]
 83%|████████▎ | 215/260 [00:53<00:21,  2.09it/s]
 83%|████████▎ | 216/260 [00:53<00:21,  2.07it/s]
 83%|████████▎ | 217/260 [00:54<00:20,  2.11it/s]
 84%|████████▍ | 218/260 [00:54<00:20,  2.08it/s]
 84%|████████▍ | 219/260 [00:55<00:19,  2.12it/s]
 85%|████████▍ | 220/260 [00:55<00:18,  2.15it/s]
 85%|████████▌ | 221/260 [00:56<00:18,  2.11it/s]
 85%|████████▌ | 222/260 [00:56<00:17,  2.12it/s]
 86%|████████▌ | 223/260 [00:57<00:17,  2.15it/s]
 86%|████████▌ | 224/260 [00:57<00:16,  2.16it/s]
 87%|████████▋ | 225/260 [00:58<00:16,  2.17it/s]
 87%|████████▋ | 226/260 [00:58<00:15,  2.18it/s]
 87%|████████▋ | 227/260 [00:59<00:15,  2.17it/s]
 88%|████████▊ | 228/260 [00:59<00:14,  2.17it/s]
 88%|████████▊ | 229/260 [00:59<00:14,  2.17it/s]
 88%|████████▊ | 230/260 [01:00<00:13,  2.17it/s]
 89%|████████▉ | 231/260 [01:00<00:13,  2.19it/s]
 89%|████████▉ | 232/260 [01:01<00:12,  2.19it/s]
 90%|████████▉ | 233/260 [01:01<00:12,  2.18it/s]
 90%|█████████ | 234/260 [01:02<00:11,  2.17it/s]
 90%|█████████ | 235/260 [01:02<00:11,  2.16it/s]
 91%|█████████ | 236/260 [01:03<00:11,  2.16it/s]
 91%|█████████ | 237/260 [01:03<00:10,  2.15it/s]
 92%|█████████▏| 238/260 [01:04<00:10,  2.15it/s]
 92%|█████████▏| 239/260 [01:04<00:09,  2.18it/s]
 92%|█████████▏| 240/260 [01:05<00:09,  2.20it/s]
 93%|█████████▎| 241/260 [01:05<00:08,  2.17it/s]
 93%|█████████▎| 242/260 [01:05<00:08,  2.16it/s]
 93%|█████████▎| 243/260 [01:06<00:07,  2.15it/s]
 94%|█████████▍| 244/260 [01:06<00:07,  2.20it/s]
 94%|█████████▍| 245/260 [01:07<00:06,  2.17it/s]
 95%|█████████▍| 246/260 [01:07<00:06,  2.15it/s]
 95%|█████████▌| 247/260 [01:08<00:06,  2.14it/s]
 95%|█████████▌| 248/260 [01:08<00:05,  2.19it/s]
 96%|█████████▌| 249/260 [01:09<00:05,  2.15it/s]
 96%|█████████▌| 250/260 [01:09<00:04,  2.13it/s]
 97%|█████████▋| 251/260 [01:10<00:04,  2.18it/s]
 97%|█████████▋| 252/260 [01:10<00:03,  2.15it/s]
 97%|█████████▋| 253/260 [01:11<00:03,  2.13it/s]
 98%|█████████▊| 254/260 [01:11<00:02,  2.18it/s]
 98%|████���████▊| 255/260 [01:12<00:02,  2.14it/s]
 98%|█████████▊| 256/260 [01:12<00:01,  2.17it/s]
 99%|█████████▉| 257/260 [01:12<00:01,  2.14it/s]
 99%|█████████▉| 258/260 [01:13<00:00,  2.11it/s]
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_last2_progressive_gate_20260730/progressive_confidence/checkpoint-500",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "A boy sits on a stoop while playing a flute for an audience of kittens on the sidewalk.",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20270027,
+  "generation_seconds": 74.42293643951416,
+  "total_seconds": 92.43578934669495,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": false,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0032.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0064.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0096.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0128.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0192.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0224.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0256.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/confidence/history_frames/step_0259.png"
+  ]
+}

omni_diffusion/mechanism_cases/boy_kittens/dprm/omni_t2i_progressive_confidence.json

@@ -0,0 +1,122 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_last2_progressive_gate_20260730/progressive_confidence/checkpoint-500",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "A boy sits on a stoop while playing a flute for an audience of kittens on the sidewalk.",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20270027,
+  "generation_seconds": 74.14197134971619,
+  "total_seconds": 92.23910331726074,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": false,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0032.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0064.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0096.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0128.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0192.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0224.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0256.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0259.png"
+  ],
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.3,
+        "requested_rank_bin": null,
+        "candidate_count": 164,
+        "candidate_index": 103,
+        "sequence_position": 231,
+        "visual_index": 192,
+        "confidence": 0.02623593993484974,
+        "raw_order_score": -3.640625,
+        "confidence_bin": 0,
+        "rank_quantile": 0.30061349272727966,
+        "rank_bin": 2,
+        "aux_bin": 12,
+        "phase": 2,
+        "provisional_token_id": 174471,
+        "default_candidate_indices": [
+          98
+        ],
+        "default_candidate_index": 98,
+        "default_sequence_position": 225,
+        "default_visual_index": 186,
+        "default_confidence": 0.10215643048286438,
+        "default_raw_order_score": -2.28125,
+        "default_rank_quantile": 1.0,
+        "confidence_gap_from_default": -0.07592049241065979,
+        "overridden_candidate_indices": [
+          103
+        ],
+        "row": 12,
+        "column": 0,
+        "row_normalized": 0.8,
+        "column_normalized": 0.0,
+        "center_distance": 8.74642784226795,
+        "local_neighbor_count": 5,
+        "local_revealed_neighbor_count": 0,
+        "local_revealed_fraction": 0.0,
+        "nearest_revealed_manhattan": 2
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.3,
+    "requested_rank_bin": null,
+    "candidate_count": 164,
+    "candidate_index": 103,
+    "sequence_position": 231,
+    "visual_index": 192,
+    "confidence": 0.02623593993484974,
+    "raw_order_score": -3.640625,
+    "confidence_bin": 0,
+    "rank_quantile": 0.30061349272727966,
+    "rank_bin": 2,
+    "aux_bin": 12,
+    "phase": 2,
+    "provisional_token_id": 174471,
+    "default_candidate_indices": [
+      98
+    ],
+    "default_candidate_index": 98,
+    "default_sequence_position": 225,
+    "default_visual_index": 186,
+    "default_confidence": 0.10215643048286438,
+    "default_raw_order_score": -2.28125,
+    "default_rank_quantile": 1.0,
+    "confidence_gap_from_default": -0.07592049241065979,
+    "overridden_candidate_indices": [
+      103
+    ],
+    "row": 12,
+    "column": 0,
+    "row_normalized": 0.8,
+    "column_normalized": 0.0,
+    "center_distance": 8.74642784226795,
+    "local_neighbor_count": 5,
+    "local_revealed_neighbor_count": 0,
+    "local_revealed_fraction": 0.0,
+    "nearest_revealed_manhattan": 2
+  }
+}

omni_diffusion/mechanism_cases/boy_kittens/dprm/run.log

@@ -0,0 +1,192 @@
+[2026-08-14 22:53:42,566] [INFO] [real_accelerator.py:222:get_accelerator] Setting ds_accelerator to cuda (auto detect)
+/home/nitanda_sub/mark/env/lib/python3.10/site-packages/pydub/utils.py:170: RuntimeWarning: Couldn't find ffmpeg or avconv - defaulting to ffmpeg, but may not work
+  warn("Couldn't find ffmpeg or avconv - defaulting to ffmpeg, but may not work", RuntimeWarning)
+WARNING:root:trust_remote_code: True
+Notice: ffmpeg is not installed. torchaudio is used to load audio
+If you want to use ffmpeg backend to load audio, please install it by:
+	sudo apt install ffmpeg # ubuntu
+	# brew install ffmpeg # mac
+Detect model requirements, begin to install it: /home/nitanda_sub/mark/.cache/huggingface/hub/models--FunAudioLLM--SenseVoiceSmall/snapshots/3847d57b6bdf2dd8875cb1508d2af43d80a16bf7/requirements.txt
+fail to install model requirements! 
+error error: externally-managed-environment
+
+× This environment is externally managed
+╰─> To install Python packages system-wide, try apt install
+    python3-xyz, where xyz is the package you are trying to
+    install.
+    
+    If you wish to install a non-Debian-packaged Python package,
+    create a virtual environment using python3 -m venv path/to/venv.
+    Then use path/to/venv/bin/python and path/to/venv/bin/pip. Make
+    sure you have python3-full installed.
+    
+    If you wish to install a non-Debian packaged Python application,
+    it may be easiest to use pipx install xyz, which will manage a
+    virtual environment for you. Make sure you have pipx installed.
+    
+    See /usr/share/doc/python3.12/README.venv for more information.
+
+note: If you believe this is a mistake, please contact your Python installation or OS distribution provider. You can override this, at the risk of breaking your Python installation or OS, by passing --break-system-packages.
+hint: See PEP 668 for the detailed specification.
+
+Loading remote code failed: model, No module named 'model'
+
+WARNING:root:trust_remote_code: True
+Detect model requirements, begin to install it: /home/nitanda_sub/mark/.cache/huggingface/hub/models--FunAudioLLM--SenseVoiceSmall/snapshots/3847d57b6bdf2dd8875cb1508d2af43d80a16bf7/requirements.txt
+fail to install model requirements! 
+error error: externally-managed-environment
+
+× This environment is externally managed
+╰─> To install Python packages system-wide, try apt install
+    python3-xyz, where xyz is the package you are trying to
+    install.
+    
+    If you wish to install a non-Debian-packaged Python package,
+    create a virtual environment using python3 -m venv path/to/venv.
+    Then use path/to/venv/bin/python and path/to/venv/bin/pip. Make
+    sure you have python3-full installed.
+    
+    If you wish to install a non-Debian packaged Python application,
+    it may be easiest to use pipx install xyz, which will manage a
+    virtual environment for you. Make sure you have pipx installed.
+    
+    See /usr/share/doc/python3.12/README.venv for more information.
+
+note: If you believe this is a mistake, please contact your Python installation or OS distribution provider. You can override this, at the risk of breaking your Python installation or OS, by passing --break-system-packages.
+hint: See PEP 668 for the detailed specification.
+
+Loading remote code failed: model, No module named 'model'
+
+INFO:omni_diffusion.tokenizer_magvitv2:self.rank=None
+INFO:omni_diffusion.tokenizer_magvitv2:self.device='cuda:0' Loading MAGVITv2
+INFO:omni_diffusion.tokenizer_magvitv2:self.device='cuda:0' Loading MAGVITv2 done
+target_ratios [(1, 1), (1, 2), (2, 1), (3, 1), (1, 3), (2, 2), (4, 1), (1, 4), (5, 1), (1, 5), (1, 6), (6, 1), (3, 2), (2, 3), (7, 1), (1, 7), (4, 2), (2, 4), (1, 8), (8, 1), (1, 9), (3, 3), (9, 1), (2, 5), (5, 2), (10, 1), (1, 10), (11, 1), (1, 11), (12, 1), (3, 4), (4, 3), (1, 12), (6, 2), (2, 6)]
+possible_resolutions [[512, 512], [512, 1024], [1024, 512], [1536, 512], [512, 1536], [1024, 1024], [2048, 512], [512, 2048], [2560, 512], [512, 2560], [512, 3072], [3072, 512], [1536, 1024], [1024, 1536], [3584, 512], [512, 3584], [2048, 1024], [1024, 2048], [512, 4096], [4096, 512], [512, 4608], [1536, 1536], [4608, 512], [1024, 2560], [2560, 1024], [5120, 512], [512, 5120], [5632, 512], [512, 5632], [6144, 512], [1536, 2048], [2048, 1536], [512, 6144], [3072, 1024], [1024, 3072]]
+Working with z of shape (1, 13, 16, 16) = 3328 dimensions.
+Look-up free quantizer with codebook size: 8192
+Working with z of shape (1, 13, 16, 16) = 3328 dimensions.
+Look-up free quantizer with codebook size: 8192
+denoise algorithm: entropy-penalty
+
  0%|          | 0/260 [00:00<?, ?it/s]
  0%|          | 1/260 [00:00<00:54,  4.74it/s]
  1%|          | 2/260 [00:00<00:41,  6.21it/s]
  2%|▏         | 4/260 [00:00<00:28,  8.99it/s]
  2%|▏         | 6/260 [00:00<00:25, 10.15it/s]
  3%|▎         | 8/260 [00:00<00:24, 10.48it/s]
  4%|▍         | 10/260 [00:01<00:25,  9.96it/s]
  5%|▍         | 12/260 [00:01<00:24, 10.18it/s]
  5%|▌         | 14/260 [00:01<00:23, 10.34it/s]
  6%|▌         | 16/260 [00:01<00:24,  9.79it/s]
  7%|▋         | 17/260 [00:01<00:24,  9.75it/s]
  7%|▋         | 18/260 [00:01<00:24,  9.77it/s]
  8%|▊         | 20/260 [00:02<00:24,  9.81it/s]
  8%|▊         | 21/260 [00:02<00:24,  9.85it/s]
  8%|▊         | 22/260 [00:02<00:26,  8.90it/s]
  9%|▉         | 23/260 [00:02<00:26,  9.02it/s]
  9%|▉         | 24/260 [00:02<00:25,  9.09it/s]
 10%|▉         | 25/260 [00:02<00:25,  9.23it/s]
 10%|█         | 26/260 [00:02<00:25,  9.18it/s]
 10%|█         | 27/260 [00:02<00:28,  8.19it/s]
 11%|█         | 28/260 [00:03<00:27,  8.37it/s]
 11%|█         | 29/260 [00:03<00:26,  8.59it/s]
 12%|█▏        | 30/260 [00:03<00:26,  8.61it/s]
 12%|█▏        | 31/260 [00:03<00:29,  7.76it/s]
 12%|█▏        | 32/260 [00:03<00:28,  7.97it/s]
 13%|█▎        | 33/260 [00:03<00:28,  7.98it/s]
 13%|█▎        | 34/260 [00:03<00:30,  7.32it/s]
 13%|█▎        | 35/260 [00:03<00:29,  7.56it/s]
 14%|█▍        | 36/260 [00:04<00:28,  7.73it/s]
 14%|█▍        | 37/260 [00:04<00:28,  7.92it/s]
 15%|█▍        | 38/260 [00:04<00:30,  7.22it/s]
 15%|█▌        | 39/260 [00:04<00:29,  7.42it/s]
 15%|█▌        | 40/260 [00:04<00:29,  7.54it/s]
 16%|█▌        | 41/260 [00:04<00:30,  7.25it/s]
 16%|█▌        | 42/260 [00:04<00:30,  7.04it/s]
 17%|█▋        | 43/260 [00:05<00:30,  7.20it/s]
 17%|█▋        | 44/260 [00:05<00:29,  7.30it/s]
 17%|█▋        | 45/260 [00:05<00:32,  6.71it/s]
 18%|█▊        | 46/260 [00:05<00:30,  6.92it/s]
 18%|█▊        | 47/260 [00:05<00:30,  7.06it/s]
 18%|█▊        | 48/260 [00:05<00:32,  6.56it/s]
 19%|█▉        | 49/260 [00:05<00:31,  6.77it/s]
 19%|█▉        | 50/260 [00:06<00:30,  6.90it/s]
 20%|█▉        | 51/260 [00:06<00:32,  6.42it/s]
 20%|██        | 52/260 [00:06<00:31,  6.61it/s]
 20%|██        | 53/260 [00:06<00:30,  6.73it/s]
 21%|██        | 54/260 [00:06<00:32,  6.27it/s]
 21%|██        | 55/260 [00:06<00:31,  6.46it/s]
 22%|██▏       | 56/260 [00:07<00:33,  6.03it/s]
 22%|██▏       | 57/260 [00:07<00:32,  6.24it/s]
 22%|██▏       | 58/260 [00:07<00:32,  6.30it/s]
 23%|██▎       | 59/260 [00:07<00:33,  5.94it/s]
 23%|██▎       | 60/260 [00:07<00:32,  6.13it/s]
 23%|██▎       | 61/260 [00:07<00:34,  5.80it/s]
 24%|██▍       | 62/260 [00:08<00:32,  6.00it/s]
 24%|██▍       | 63/260 [00:08<00:32,  6.09it/s]
 25%|██▍       | 64/260 [00:08<00:34,  5.73it/s]
 25%|██▌       | 65/260 [00:08<00:33,  5.85it/s]
 25%|██▌       | 66/260 [00:08<00:34,  5.57it/s]
 26%|██▌       | 67/260 [00:08<00:33,  5.72it/s]
 26%|██▌       | 68/260 [00:09<00:35,  5.46it/s]
 27%|██▋       | 69/260 [00:09<00:33,  5.66it/s]
 27%|██▋       | 70/260 [00:09<00:33,  5.67it/s]
 27%|██▋       | 71/260 [00:09<00:34,  5.43it/s]
 28%|██▊       | 72/260 [00:09<00:33,  5.59it/s]
 28%|██▊       | 73/260 [00:10<00:35,  5.32it/s]
 28%|██▊       | 74/260 [00:10<00:33,  5.50it/s]
 29%|██▉       | 75/260 [00:10<00:35,  5.20it/s]
 29%|██▉       | 76/260 [00:10<00:34,  5.39it/s]
 30%|██▉       | 77/260 [00:10<00:35,  5.13it/s]
 30%|███       | 78/260 [00:10<00:34,  5.31it/s]
 30%|███       | 79/260 [00:11<00:35,  5.10it/s]
 31%|███       | 80/260 [00:11<00:34,  5.23it/s]
 31%|███       | 81/260 [00:11<00:35,  5.08it/s]
 32%|███▏      | 82/260 [00:11<00:34,  5.19it/s]
 32%|███▏      | 83/260 [00:11<00:33,  5.31it/s]
 32%|███▏      | 84/260 [00:12<00:35,  5.01it/s]
 33%|███▎      | 85/260 [00:12<00:34,  5.14it/s]
 33%|███▎      | 86/260 [00:12<00:35,  4.90it/s]
 33%|███▎      | 87/260 [00:12<00:34,  5.06it/s]
 34%|███▍      | 88/260 [00:12<00:35,  4.83it/s]
 34%|███▍      | 89/260 [00:13<00:34,  4.99it/s]
 35%|███▍      | 90/260 [00:13<00:35,  4.77it/s]
 35%|███▌      | 91/260 [00:13<00:34,  4.92it/s]
 35%|███▌      | 92/260 [00:13<00:35,  4.68it/s]
 36%|███▌      | 93/260 [00:14<00:34,  4.84it/s]
 36%|███▌      | 94/260 [00:14<00:35,  4.63it/s]
 37%|███▋      | 95/260 [00:14<00:36,  4.55it/s]
 37%|███▋      | 96/260 [00:14<00:35,  4.66it/s]
 37%|███▋      | 97/260 [00:14<00:38,  4.28it/s]
 38%|███▊      | 98/260 [00:15<00:38,  4.23it/s]
 38%|███▊      | 99/260 [00:15<00:36,  4.45it/s]
 38%|███▊      | 100/260 [00:15<00:36,  4.34it/s]
 39%|███▉      | 101/260 [00:15<00:35,  4.52it/s]
 39%|███▉      | 102/260 [00:16<00:36,  4.37it/s]
 40%|███▉      | 103/260 [00:16<00:34,  4.52it/s]
 40%|████      | 104/260 [00:16<00:35,  4.35it/s]
 40%|████      | 105/260 [00:16<00:35,  4.37it/s]
 41%|████      | 106/260 [00:17<00:35,  4.35it/s]
 41%|████      | 107/260 [00:17<00:35,  4.26it/s]
 42%|████▏     | 108/260 [00:17<00:34,  4.40it/s]
 42%|████▏     | 109/260 [00:17<00:35,  4.24it/s]
 42%|████▏     | 110/260 [00:17<00:34,  4.37it/s]
 43%|████▎     | 111/260 [00:18<00:35,  4.23it/s]
 43%|████▎     | 112/260 [00:18<00:33,  4.35it/s]
 43%|████▎     | 113/260 [00:18<00:35,  4.18it/s]
 44%|████▍     | 114/260 [00:18<00:35,  4.08it/s]
 44%|████▍     | 115/260 [00:19<00:34,  4.19it/s]
 45%|████▍     | 116/260 [00:19<00:35,  4.07it/s]
 45%|████▌     | 117/260 [00:19<00:34,  4.17it/s]
 45%|████▌     | 118/260 [00:19<00:35,  4.06it/s]
 46%|████▌     | 119/260 [00:20<00:35,  3.97it/s]
 46%|████▌     | 120/260 [00:20<00:34,  4.08it/s]
 47%|████▋     | 121/260 [00:20<00:34,  3.98it/s]
 47%|████▋     | 122/260 [00:20<00:33,  4.09it/s]
 47%|████▋     | 123/260 [00:21<00:34,  3.95it/s]
 48%|████▊     | 124/260 [00:21<00:35,  3.84it/s]
 48%|████▊     | 125/260 [00:21<00:33,  3.98it/s]
 48%|████▊     | 126/260 [00:21<00:34,  3.88it/s]
 49%|████▉     | 127/260 [00:22<00:35,  3.79it/s]
 49%|████▉     | 128/260 [00:22<00:33,  3.92it/s]
 50%|████▉     | 129/260 [00:22<00:34,  3.81it/s]
 50%|█████     | 130/260 [00:22<00:33,  3.93it/s]
 50%|█████     | 131/260 [00:23<00:32,  3.92it/s]
 51%|█████     | 132/260 [00:23<00:33,  3.79it/s]
 51%|█████     | 133/260 [00:23<00:32,  3.90it/s]
 52%|█████▏    | 134/260 [00:24<00:34,  3.70it/s]
 52%|█████▏    | 135/260 [00:24<00:34,  3.63it/s]
 52%|█████▏    | 136/260 [00:24<00:33,  3.75it/s]
 53%|█████▎    | 137/260 [00:24<00:33,  3.64it/s]
 53%|█████▎    | 138/260 [00:25<00:34,  3.57it/s]
 53%|█████▎    | 139/260 [00:25<00:32,  3.70it/s]
 54%|█████▍    | 140/260 [00:25<00:33,  3.60it/s]
 54%|█████▍    | 141/260 [00:26<00:33,  3.55it/s]
 55%|█████▍    | 142/260 [00:26<00:32,  3.65it/s]
 55%|█████▌    | 143/260 [00:26<00:32,  3.58it/s]
 55%|█████▌    | 144/260 [00:26<00:33,  3.51it/s]
 56%|█████▌    | 145/260 [00:27<00:31,  3.62it/s]
 56%|█████▌    | 146/260 [00:27<00:32,  3.52it/s]
 57%|█████▋    | 147/260 [00:27<00:32,  3.45it/s]
 57%|█████▋    | 148/260 [00:27<00:31,  3.56it/s]
 57%|█████▋    | 149/260 [00:28<00:31,  3.47it/s]
 58%|█████▊    | 150/260 [00:28<00:32,  3.41it/s]
 58%|█████▊    | 151/260 [00:28<00:31,  3.50it/s]
 58%|█████▊    | 152/260 [00:29<00:31,  3.43it/s]
 59%|█████▉    | 153/260 [00:29<00:31,  3.37it/s]
 59%|█████▉    | 154/260 [00:29<00:30,  3.45it/s]
 60%|█████▉    | 155/260 [00:30<00:31,  3.37it/s]
 60%|██████    | 156/260 [00:30<00:31,  3.31it/s]
 60%|██████    | 157/260 [00:30<00:31,  3.27it/s]
 61%|██████    | 158/260 [00:30<00:30,  3.37it/s]
 61%|██████    | 159/260 [00:31<00:30,  3.32it/s]
 62%|██████▏   | 160/260 [00:31<00:30,  3.26it/s]
 62%|██████▏   | 161/260 [00:31<00:29,  3.39it/s]
 62%|██████▏   | 162/260 [00:32<00:29,  3.30it/s]
 63%|██████▎   | 163/260 [00:32<00:29,  3.26it/s]
 63%|██████▎   | 164/260 [00:32<00:29,  3.22it/s]
 63%|██████▎   | 165/260 [00:33<00:28,  3.34it/s]
 64%|██████▍   | 166/260 [00:33<00:28,  3.28it/s]
 64%|██████▍   | 167/260 [00:33<00:28,  3.23it/s]
 65%|██████▍   | 168/260 [00:34<00:27,  3.30it/s]
 65%|██████▌   | 169/260 [00:34<00:28,  3.24it/s]
 65%|██████▌   | 170/260 [00:34<00:28,  3.19it/s]
 66%|██████▌   | 171/260 [00:34<00:28,  3.15it/s]
 66%|██████▌   | 172/260 [00:35<00:27,  3.26it/s]
 67%|██████▋   | 173/260 [00:35<00:27,  3.19it/s]
 67%|██████▋   | 174/260 [00:35<00:27,  3.14it/s]
 67%|██████▋   | 175/260 [00:36<00:27,  3.14it/s]
 68%|██████▊   | 176/260 [00:36<00:26,  3.20it/s]
 68%|██████▊   | 177/260 [00:36<00:26,  3.15it/s]
 68%|██████▊   | 178/260 [00:37<00:26,  3.11it/s]
 69%|██████▉   | 179/260 [00:37<00:25,  3.18it/s]
 69%|██████▉   | 180/260 [00:37<00:25,  3.14it/s]
 70%|██████▉   | 181/260 [00:38<00:25,  3.08it/s]
 70%|███████   | 182/260 [00:38<00:25,  3.04it/s]
 70%|███████   | 183/260 [00:38<00:24,  3.12it/s]
 71%|███████   | 184/260 [00:39<00:24,  3.06it/s]
 71%|███████   | 185/260 [00:39<00:24,  3.01it/s]
 72%|███████▏  | 186/260 [00:39<00:24,  2.98it/s]
 72%|███████▏  | 187/260 [00:40<00:23,  3.07it/s]
 72%|███████▏  | 188/260 [00:40<00:23,  3.02it/s]
 73%|███████▎  | 189/260 [00:40<00:23,  2.97it/s]
 73%|███████▎  | 190/260 [00:41<00:23,  2.94it/s]
 73%|███████▎  | 191/260 [00:41<00:22,  3.01it/s]
 74%|███████▍  | 192/260 [00:41<00:22,  2.96it/s]
 74%|███████▍  | 193/260 [00:42<00:22,  2.92it/s]
 75%|███████▍  | 194/260 [00:42<00:21,  3.00it/s]
 75%|███████▌  | 195/260 [00:42<00:22,  2.88it/s]
 75%|███████▌  | 196/260 [00:43<00:28,  2.22it/s]
 76%|███████▌  | 197/260 [00:44<00:32,  1.96it/s]
 76%|███████▌  | 198/260 [00:44<00:33,  1.87it/s]
 77%|███████▋  | 199/260 [00:45<00:33,  1.80it/s]
 77%|███████▋  | 200/260 [00:45<00:33,  1.81it/s]
 77%|███████▋  | 201/260 [00:46<00:32,  1.81it/s]
 78%|███████▊  | 202/260 [00:47<00:31,  1.86it/s]
 78%|███████▊  | 203/260 [00:47<00:30,  1.86it/s]
 78%|███████▊  | 204/260 [00:48<00:29,  1.92it/s]
 79%|███████▉  | 205/260 [00:48<00:28,  1.95it/s]
 79%|███████▉  | 206/260 [00:49<00:27,  1.97it/s]
 80%|███████▉  | 207/260 [00:49<00:26,  1.99it/s]
 80%|████████  | 208/260 [00:50<00:25,  2.01it/s]
 80%|████████  | 209/260 [00:50<00:25,  2.01it/s]
 81%|████████  | 210/260 [00:51<00:24,  2.02it/s]
 81%|████████  | 211/260 [00:51<00:23,  2.09it/s]
 82%|████████▏ | 212/260 [00:51<00:23,  2.07it/s]
 82%|████████▏ | 213/260 [00:52<00:22,  2.12it/s]
 82%|████████▏ | 214/260 [00:52<00:22,  2.09it/s]
 83%|████████▎ | 215/260 [00:53<00:21,  2.14it/s]
 83%|████████▎ | 216/260 [00:53<00:20,  2.10it/s]
 83%|████████▎ | 217/260 [00:54<00:20,  2.13it/s]
 84%|████████▍ | 218/260 [00:54<00:19,  2.16it/s]
 84%|████████▍ | 219/260 [00:55<00:18,  2.17it/s]
 85%|████████▍ | 220/260 [00:55<00:18,  2.12it/s]
 85%|████████▌ | 221/260 [00:56<00:18,  2.15it/s]
 85%|████████▌ | 222/260 [00:56<00:17,  2.16it/s]
 86%|████████▌ | 223/260 [00:57<00:16,  2.18it/s]
 86%|████████▌ | 224/260 [00:57<00:16,  2.18it/s]
 87%|████████▋ | 225/260 [00:57<00:16,  2.18it/s]
 87%|████████▋ | 226/260 [00:58<00:15,  2.19it/s]
 87%|████████▋ | 227/260 [00:58<00:15,  2.19it/s]
 88%|████████▊ | 228/260 [00:59<00:14,  2.18it/s]
 88%|████████▊ | 229/260 [00:59<00:14,  2.18it/s]
 88%|████████▊ | 230/260 [01:00<00:13,  2.18it/s]
 89%|████████▉ | 231/260 [01:00<00:13,  2.17it/s]
 89%|████████▉ | 232/260 [01:01<00:12,  2.17it/s]
 90%|████████▉ | 233/260 [01:01<00:12,  2.16it/s]
 90%|█████████ | 234/260 [01:02<00:12,  2.16it/s]
 90%|█████████ | 235/260 [01:02<00:11,  2.15it/s]
 91%|█████████ | 236/260 [01:02<00:10,  2.22it/s]
 91%|█████████ | 237/260 [01:03<00:10,  2.20it/s]
 92%|█████████▏| 238/260 [01:03<00:10,  2.17it/s]
 92%|█████████▏| 239/260 [01:04<00:09,  2.16it/s]
 92%|█████████▏| 240/260 [01:04<00:09,  2.15it/s]
 93%|█████████▎| 241/260 [01:05<00:08,  2.21it/s]
 93%|█████████▎| 242/260 [01:05<00:08,  2.18it/s]
 93%|█████████▎| 243/260 [01:06<00:07,  2.16it/s]
 94%|█████████▍| 244/260 [01:06<00:07,  2.14it/s]
 94%|█████████▍| 245/260 [01:07<00:06,  2.20it/s]
 95%|█████████▍| 246/260 [01:07<00:06,  2.16it/s]
 95%|█████████▌| 247/260 [01:08<00:06,  2.15it/s]
 95%|█████████▌| 248/260 [01:08<00:05,  2.20it/s]
 96%|█████████▌| 249/260 [01:08<00:05,  2.16it/s]
 96%|█████████▌| 250/260 [01:09<00:04,  2.25it/s]
 97%|█████████▋| 251/260 [01:09<00:04,  2.19it/s]
 97%|█████████▋| 252/260 [01:10<00:03,  2.15it/s]
 97%|█████████▋| 253/260 [01:10<00:03,  2.19it/s]
 98%|█████████▊| 254/260 [01:11<00:02,  2.14it/s]
 98%|█████████▊| 255/260 [01:11<00:02,  2.16it/s]
 98%|█████████▊| 256/260 [01:12<00:01,  2.14it/s]
 99%|█████████▉| 257/260 [01:12<00:01,  2.11it/s]
 99%|█████████▉| 258/260 [01:13<00:00,  2.15it/s]
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_last2_progressive_gate_20260730/progressive_confidence/checkpoint-500",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "A boy sits on a stoop while playing a flute for an audience of kittens on the sidewalk.",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20270027,
+  "generation_seconds": 74.14197134971619,
+  "total_seconds": 92.23910331726074,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": false,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0032.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0064.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0096.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0128.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0192.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0224.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0256.png",
+    "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_boy_kittens_intermediate_20260814/dprm/history_frames/step_0259.png"
+  ],
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.3,
+        "requested_rank_bin": null,
+        "candidate_count": 164,
+        "candidate_index": 103,
+        "sequence_position": 231,
+        "visual_index": 192,
+        "confidence": 0.02623593993484974,
+        "raw_order_score": -3.640625,
+        "confidence_bin": 0,
+        "rank_quantile": 0.30061349272727966,
+        "rank_bin": 2,
+        "aux_bin": 12,
+        "phase": 2,
+        "provisional_token_id": 174471,
+        "default_candidate_indices": [
+          98
+        ],
+        "default_candidate_index": 98,
+        "default_sequence_position": 225,
+        "default_visual_index": 186,
+        "default_confidence": 0.10215643048286438,
+        "default_raw_order_score": -2.28125,
+        "default_rank_quantile": 1.0,
+        "confidence_gap_from_default": -0.07592049241065979,
+        "overridden_candidate_indices": [
+          103
+        ],
+        "row": 12,
+        "column": 0,
+        "row_normalized": 0.8,
+        "column_normalized": 0.0,
+        "center_distance": 8.74642784226795,
+        "local_neighbor_count": 5,
+        "local_revealed_neighbor_count": 0,
+        "local_revealed_fraction": 0.0,
+        "nearest_revealed_manhattan": 2
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.3,
+    "requested_rank_bin": null,
+    "candidate_count": 164,
+    "candidate_index": 103,
+    "sequence_position": 231,
+    "visual_index": 192,
+    "confidence": 0.02623593993484974,
+    "raw_order_score": -3.640625,
+    "confidence_bin": 0,
+    "rank_quantile": 0.30061349272727966,
+    "rank_bin": 2,
+    "aux_bin": 12,
+    "phase": 2,
+    "provisional_token_id": 174471,
+    "default_candidate_indices": [
+      98
+    ],
+    "default_candidate_index": 98,
+    "default_sequence_position": 225,
+    "default_visual_index": 186,
+    "default_confidence": 0.10215643048286438,
+    "default_raw_order_score": -2.28125,
+    "default_rank_quantile": 1.0,
+    "confidence_gap_from_default": -0.07592049241065979,
+    "overridden_candidate_indices": [
+      103
+    ],
+    "row": 12,
+    "column": 0,
+    "row_normalized": 0.8,
+    "column_normalized": 0.0,
+    "center_distance": 8.74642784226795,
+    "local_neighbor_count": 5,
+    "local_revealed_neighbor_count": 0,
+    "local_revealed_fraction": 0.0,
+    "nearest_revealed_manhattan": 2
+  }
+}

omni_diffusion/mechanism_cases/manifest.json

@@ -0,0 +1,48 @@
+{
+  "schema_version": 1,
+  "role": "fixed mechanism diagnostic; excluded from the 512-prompt confirmation mean",
+  "checkpoint": "Omni confidence checkpoint-500",
+  "action_step": 96,
+  "candidate_rank_quantiles": [0.15, 0.30, 0.70, 0.85],
+  "selection": "hard terminal-utility tilt with confidence fallback",
+  "terminal_utility": "CLIP-L/14 cosine + 0.01 * LAION aesthetic score",
+  "cases": [
+    {
+      "id": "beach",
+      "prompt_id": "20270085",
+      "seed": 20270085,
+      "prompt": "Two little boys and a little girl playing on a sunny beach with clear blue skies, as they explore the shore and dig in the sand with their toy pail and shovel.",
+      "selected_quantile": 0.15,
+      "default_visual_index": 156,
+      "selected_visual_index": 229,
+      "default_action_confidence": 0.03164658322930336,
+      "selected_action_confidence": 0.020754337310791016,
+      "confidence_clip_l14": 0.27711355686187744,
+      "dprm_clip_l14": 0.3314807415008545,
+      "confidence_clip_b32": 0.32946068048477173,
+      "dprm_clip_b32": 0.34731602668762207,
+      "hf_confidence_dir": "omni_diffusion/mechanism_cases/beach/confidence",
+      "hf_dprm_dir": "omni_diffusion/mechanism_cases/beach/dprm"
+    },
+    {
+      "id": "boy_kittens",
+      "prompt_id": "20270027",
+      "seed": 20270027,
+      "prompt": "A boy sits on a stoop while playing a flute for an audience of kittens on the sidewalk.",
+      "selected_quantile": 0.30,
+      "default_visual_index": 186,
+      "selected_visual_index": 192,
+      "default_action_confidence": 0.10215643048286438,
+      "selected_action_confidence": 0.02623593993484974,
+      "confidence_clip_l14": 0.28843873739242554,
+      "dprm_clip_l14": 0.31169044971466064,
+      "confidence_clip_b32": 0.30447253584861755,
+      "dprm_clip_b32": 0.3159780502319336,
+      "hf_confidence_dir": "omni_diffusion/mechanism_cases/boy_kittens/confidence",
+      "hf_dprm_dir": "omni_diffusion/mechanism_cases/boy_kittens/dprm"
+    }
+  ],
+  "records": "omni_diffusion/mechanism_cases/audit_records_b32.json",
+  "renderer": "integrations/omni_diffusion/matched/scripts/render_omni_mechanism_cases.py",
+  "paper_figure_sha256": "4e0316c616b9a461b8fc49177670b4a8f0cda2d5253d2f600aecb715f95bd405"
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1154/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "a snowy owl standing in a grassy field",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281154,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 21.760518074035645,
+  "total_seconds": 21.79355812072754,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1154/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 38,
+        "sequence_position": 124,
+        "visual_index": 96,
+        "confidence": 0.43346792459487915,
+        "raw_order_score": -0.8359375,
+        "confidence_bin": 3,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 4,
+        "phase": 2,
+        "provisional_token_id": 170436,
+        "default_candidate_indices": [
+          9
+        ],
+        "default_candidate_index": 9,
+        "default_sequence_position": 63,
+        "default_visual_index": 35,
+        "default_confidence": 0.9999989867210388,
+        "default_raw_order_score": 0.0,
+        "default_rank_quantile": 0.9748427271842957,
+        "confidence_gap_from_default": -0.5665310621261597,
+        "overridden_candidate_indices": [
+          38
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 6,
+        "column": 0,
+        "row_normalized": 0.4,
+        "column_normalized": 0.0,
+        "center_distance": 7.648529270389178,
+        "local_neighbor_count": 5,
+        "local_revealed_neighbor_count": 3,
+        "local_revealed_fraction": 0.6,
+        "nearest_revealed_manhattan": 1
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 38,
+    "sequence_position": 124,
+    "visual_index": 96,
+    "confidence": 0.43346792459487915,
+    "raw_order_score": -0.8359375,
+    "confidence_bin": 3,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 4,
+    "phase": 2,
+    "provisional_token_id": 170436,
+    "default_candidate_indices": [
+      9
+    ],
+    "default_candidate_index": 9,
+    "default_sequence_position": 63,
+    "default_visual_index": 35,
+    "default_confidence": 0.9999989867210388,
+    "default_raw_order_score": 0.0,
+    "default_rank_quantile": 0.9748427271842957,
+    "confidence_gap_from_default": -0.5665310621261597,
+    "overridden_candidate_indices": [
+      38
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 6,
+    "column": 0,
+    "row_normalized": 0.4,
+    "column_normalized": 0.0,
+    "center_distance": 7.648529270389178,
+    "local_neighbor_count": 5,
+    "local_revealed_neighbor_count": 3,
+    "local_revealed_fraction": 0.6,
+    "nearest_revealed_manhattan": 1
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1166/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "a handsaw on a table",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281166,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 21.238081693649292,
+  "total_seconds": 21.269418716430664,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1166/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 142,
+        "sequence_position": 247,
+        "visual_index": 222,
+        "confidence": 0.14633232355117798,
+        "raw_order_score": -1.921875,
+        "confidence_bin": 1,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 15,
+        "phase": 2,
+        "provisional_token_id": 172343,
+        "default_candidate_indices": [
+          152
+        ],
+        "default_candidate_index": 152,
+        "default_sequence_position": 270,
+        "default_visual_index": 245,
+        "default_confidence": 0.5394595265388489,
+        "default_raw_order_score": -0.6171875,
+        "default_rank_quantile": 0.9999999403953552,
+        "confidence_gap_from_default": -0.3931272029876709,
+        "overridden_candidate_indices": [
+          142
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 13,
+        "column": 14,
+        "row_normalized": 0.8666666666666667,
+        "column_normalized": 0.9333333333333333,
+        "center_distance": 8.514693182963201,
+        "local_neighbor_count": 8,
+        "local_revealed_neighbor_count": 3,
+        "local_revealed_fraction": 0.375,
+        "nearest_revealed_manhattan": 1
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 142,
+    "sequence_position": 247,
+    "visual_index": 222,
+    "confidence": 0.14633232355117798,
+    "raw_order_score": -1.921875,
+    "confidence_bin": 1,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 15,
+    "phase": 2,
+    "provisional_token_id": 172343,
+    "default_candidate_indices": [
+      152
+    ],
+    "default_candidate_index": 152,
+    "default_sequence_position": 270,
+    "default_visual_index": 245,
+    "default_confidence": 0.5394595265388489,
+    "default_raw_order_score": -0.6171875,
+    "default_rank_quantile": 0.9999999403953552,
+    "confidence_gap_from_default": -0.3931272029876709,
+    "overridden_candidate_indices": [
+      142
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 13,
+    "column": 14,
+    "row_normalized": 0.8666666666666667,
+    "column_normalized": 0.9333333333333333,
+    "center_distance": 8.514693182963201,
+    "local_neighbor_count": 8,
+    "local_revealed_neighbor_count": 3,
+    "local_revealed_fraction": 0.375,
+    "nearest_revealed_manhattan": 1
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1168/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "a plate with white rice topped by cooked vegetables",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281168,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 20.75342607498169,
+  "total_seconds": 20.785287141799927,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1168/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 55,
+        "sequence_position": 155,
+        "visual_index": 127,
+        "confidence": 0.4873610734939575,
+        "raw_order_score": -0.71875,
+        "confidence_bin": 3,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 7,
+        "phase": 2,
+        "provisional_token_id": 173700,
+        "default_candidate_indices": [
+          108
+        ],
+        "default_candidate_index": 108,
+        "default_sequence_position": 228,
+        "default_visual_index": 200,
+        "default_confidence": 0.9999989867210388,
+        "default_raw_order_score": 0.0,
+        "default_rank_quantile": 0.9685534238815308,
+        "confidence_gap_from_default": -0.5126379132270813,
+        "overridden_candidate_indices": [
+          55
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 7,
+        "column": 15,
+        "row_normalized": 0.4666666666666667,
+        "column_normalized": 1.0,
+        "center_distance": 7.516648189186454,
+        "local_neighbor_count": 5,
+        "local_revealed_neighbor_count": 0,
+        "local_revealed_fraction": 0.0,
+        "nearest_revealed_manhattan": 3
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 55,
+    "sequence_position": 155,
+    "visual_index": 127,
+    "confidence": 0.4873610734939575,
+    "raw_order_score": -0.71875,
+    "confidence_bin": 3,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 7,
+    "phase": 2,
+    "provisional_token_id": 173700,
+    "default_candidate_indices": [
+      108
+    ],
+    "default_candidate_index": 108,
+    "default_sequence_position": 228,
+    "default_visual_index": 200,
+    "default_confidence": 0.9999989867210388,
+    "default_raw_order_score": 0.0,
+    "default_rank_quantile": 0.9685534238815308,
+    "confidence_gap_from_default": -0.5126379132270813,
+    "overridden_candidate_indices": [
+      55
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 7,
+    "column": 15,
+    "row_normalized": 0.4666666666666667,
+    "column_normalized": 1.0,
+    "center_distance": 7.516648189186454,
+    "local_neighbor_count": 5,
+    "local_revealed_neighbor_count": 0,
+    "local_revealed_fraction": 0.0,
+    "nearest_revealed_manhattan": 3
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1172/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "A smiling sloth wearing a bowtie and holding a quarterstaff and a big book. A shiny VW van parked on grass.",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281172,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 20.9438796043396,
+  "total_seconds": 20.973559617996216,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": ",<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1172/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 34,
+        "sequence_position": 118,
+        "visual_index": 73,
+        "confidence": 0.302609384059906,
+        "raw_order_score": -1.1953125,
+        "confidence_bin": 2,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 6,
+        "phase": 2,
+        "provisional_token_id": 170143,
+        "default_candidate_indices": [
+          130
+        ],
+        "default_candidate_index": 130,
+        "default_sequence_position": 263,
+        "default_visual_index": 218,
+        "default_confidence": 0.7301880717277527,
+        "default_raw_order_score": -0.314453125,
+        "default_rank_quantile": 0.9999999403953552,
+        "confidence_gap_from_default": -0.4275786876678467,
+        "overridden_candidate_indices": [
+          34
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 4,
+        "column": 9,
+        "row_normalized": 0.26666666666666666,
+        "column_normalized": 0.6,
+        "center_distance": 3.8078865529319543,
+        "local_neighbor_count": 8,
+        "local_revealed_neighbor_count": 5,
+        "local_revealed_fraction": 0.625,
+        "nearest_revealed_manhattan": 1
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 34,
+    "sequence_position": 118,
+    "visual_index": 73,
+    "confidence": 0.302609384059906,
+    "raw_order_score": -1.1953125,
+    "confidence_bin": 2,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 6,
+    "phase": 2,
+    "provisional_token_id": 170143,
+    "default_candidate_indices": [
+      130
+    ],
+    "default_candidate_index": 130,
+    "default_sequence_position": 263,
+    "default_visual_index": 218,
+    "default_confidence": 0.7301880717277527,
+    "default_raw_order_score": -0.314453125,
+    "default_rank_quantile": 0.9999999403953552,
+    "confidence_gap_from_default": -0.4275786876678467,
+    "overridden_candidate_indices": [
+      34
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 4,
+    "column": 9,
+    "row_normalized": 0.26666666666666666,
+    "column_normalized": 0.6,
+    "center_distance": 3.8078865529319543,
+    "local_neighbor_count": 8,
+    "local_revealed_neighbor_count": 5,
+    "local_revealed_fraction": 0.625,
+    "nearest_revealed_manhattan": 1
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1174/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "three small yellow boxes",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281174,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 21.821401596069336,
+  "total_seconds": 21.850460529327393,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1174/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 42,
+        "sequence_position": 137,
+        "visual_index": 114,
+        "confidence": 0.4351644814014435,
+        "raw_order_score": -0.83203125,
+        "confidence_bin": 3,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 4,
+        "phase": 2,
+        "provisional_token_id": 173316,
+        "default_candidate_indices": [
+          114
+        ],
+        "default_candidate_index": 114,
+        "default_sequence_position": 231,
+        "default_visual_index": 208,
+        "default_confidence": 0.9999989867210388,
+        "default_raw_order_score": 0.0,
+        "default_rank_quantile": 0.9685534238815308,
+        "confidence_gap_from_default": -0.564834475517273,
+        "overridden_candidate_indices": [
+          42
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 7,
+        "column": 2,
+        "row_normalized": 0.4666666666666667,
+        "column_normalized": 0.13333333333333333,
+        "center_distance": 5.522680508593631,
+        "local_neighbor_count": 8,
+        "local_revealed_neighbor_count": 5,
+        "local_revealed_fraction": 0.625,
+        "nearest_revealed_manhattan": 1
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 42,
+    "sequence_position": 137,
+    "visual_index": 114,
+    "confidence": 0.4351644814014435,
+    "raw_order_score": -0.83203125,
+    "confidence_bin": 3,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 4,
+    "phase": 2,
+    "provisional_token_id": 173316,
+    "default_candidate_indices": [
+      114
+    ],
+    "default_candidate_index": 114,
+    "default_sequence_position": 231,
+    "default_visual_index": 208,
+    "default_confidence": 0.9999989867210388,
+    "default_raw_order_score": 0.0,
+    "default_rank_quantile": 0.9685534238815308,
+    "confidence_gap_from_default": -0.564834475517273,
+    "overridden_candidate_indices": [
+      42
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 7,
+    "column": 2,
+    "row_normalized": 0.4666666666666667,
+    "column_normalized": 0.13333333333333333,
+    "center_distance": 5.522680508593631,
+    "local_neighbor_count": 8,
+    "local_revealed_neighbor_count": 5,
+    "local_revealed_fraction": 0.625,
+    "nearest_revealed_manhattan": 1
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1175/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "the finale of a fireworks display",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281175,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 20.742891311645508,
+  "total_seconds": 20.774680614471436,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|><|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1175/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 85,
+        "sequence_position": 186,
+        "visual_index": 161,
+        "confidence": 0.1291378289461136,
+        "raw_order_score": -2.046875,
+        "confidence_bin": 1,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 8,
+        "phase": 2,
+        "provisional_token_id": 169894,
+        "default_candidate_indices": [
+          26
+        ],
+        "default_candidate_index": 26,
+        "default_sequence_position": 79,
+        "default_visual_index": 54,
+        "default_confidence": 0.5228621363639832,
+        "default_raw_order_score": -0.6484375,
+        "default_rank_quantile": 0.9999999403953552,
+        "confidence_gap_from_default": -0.39372432231903076,
+        "overridden_candidate_indices": [
+          85
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 10,
+        "column": 1,
+        "row_normalized": 0.6666666666666666,
+        "column_normalized": 0.06666666666666667,
+        "center_distance": 6.96419413859206,
+        "local_neighbor_count": 8,
+        "local_revealed_neighbor_count": 4,
+        "local_revealed_fraction": 0.5,
+        "nearest_revealed_manhattan": 1
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 85,
+    "sequence_position": 186,
+    "visual_index": 161,
+    "confidence": 0.1291378289461136,
+    "raw_order_score": -2.046875,
+    "confidence_bin": 1,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 8,
+    "phase": 2,
+    "provisional_token_id": 169894,
+    "default_candidate_indices": [
+      26
+    ],
+    "default_candidate_index": 26,
+    "default_sequence_position": 79,
+    "default_visual_index": 54,
+    "default_confidence": 0.5228621363639832,
+    "default_raw_order_score": -0.6484375,
+    "default_rank_quantile": 0.9999999403953552,
+    "confidence_gap_from_default": -0.39372432231903076,
+    "overridden_candidate_indices": [
+      85
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 10,
+    "column": 1,
+    "row_normalized": 0.6666666666666666,
+    "column_normalized": 0.06666666666666667,
+    "center_distance": 6.96419413859206,
+    "local_neighbor_count": 8,
+    "local_revealed_neighbor_count": 4,
+    "local_revealed_fraction": 0.5,
+    "nearest_revealed_manhattan": 1
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1185/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "a pair of headphones on a statue of a horse",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281185,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 21.691670179367065,
+  "total_seconds": 21.732835292816162,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|><|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1185/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 4,
+        "sequence_position": 41,
+        "visual_index": 12,
+        "confidence": 0.30978554487228394,
+        "raw_order_score": -1.171875,
+        "confidence_bin": 2,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 3,
+        "phase": 2,
+        "provisional_token_id": 175286,
+        "default_candidate_indices": [
+          84
+        ],
+        "default_candidate_index": 84,
+        "default_sequence_position": 199,
+        "default_visual_index": 170,
+        "default_confidence": 0.9999989867210388,
+        "default_raw_order_score": 0.0,
+        "default_rank_quantile": 0.9937106370925903,
+        "confidence_gap_from_default": -0.6902134418487549,
+        "overridden_candidate_indices": [
+          4
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 0,
+        "column": 12,
+        "row_normalized": 0.0,
+        "column_normalized": 0.8,
+        "center_distance": 8.74642784226795,
+        "local_neighbor_count": 5,
+        "local_revealed_neighbor_count": 3,
+        "local_revealed_fraction": 0.6,
+        "nearest_revealed_manhattan": 1
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 4,
+    "sequence_position": 41,
+    "visual_index": 12,
+    "confidence": 0.30978554487228394,
+    "raw_order_score": -1.171875,
+    "confidence_bin": 2,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 3,
+    "phase": 2,
+    "provisional_token_id": 175286,
+    "default_candidate_indices": [
+      84
+    ],
+    "default_candidate_index": 84,
+    "default_sequence_position": 199,
+    "default_visual_index": 170,
+    "default_confidence": 0.9999989867210388,
+    "default_raw_order_score": 0.0,
+    "default_rank_quantile": 0.9937106370925903,
+    "confidence_gap_from_default": -0.6902134418487549,
+    "overridden_candidate_indices": [
+      4
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 0,
+    "column": 12,
+    "row_normalized": 0.0,
+    "column_normalized": 0.8,
+    "center_distance": 8.74642784226795,
+    "local_neighbor_count": 5,
+    "local_revealed_neighbor_count": 3,
+    "local_revealed_fraction": 0.6,
+    "nearest_revealed_manhattan": 1
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1189/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "We painted the wall with cracks.",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281189,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 22.342575550079346,
+  "total_seconds": 22.37885808944702,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|><|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1189/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 19,
+        "sequence_position": 76,
+        "visual_index": 50,
+        "confidence": 0.41523683071136475,
+        "raw_order_score": -0.87890625,
+        "confidence_bin": 3,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 0,
+        "phase": 2,
+        "provisional_token_id": 175391,
+        "default_candidate_indices": [
+          68
+        ],
+        "default_candidate_index": 68,
+        "default_sequence_position": 167,
+        "default_visual_index": 141,
+        "default_confidence": 0.9999989867210388,
+        "default_raw_order_score": 0.0,
+        "default_rank_quantile": 0.9937106370925903,
+        "confidence_gap_from_default": -0.5847621560096741,
+        "overridden_candidate_indices": [
+          19
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 3,
+        "column": 2,
+        "row_normalized": 0.2,
+        "column_normalized": 0.13333333333333333,
+        "center_distance": 7.106335201775948,
+        "local_neighbor_count": 8,
+        "local_revealed_neighbor_count": 3,
+        "local_revealed_fraction": 0.375,
+        "nearest_revealed_manhattan": 1
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 19,
+    "sequence_position": 76,
+    "visual_index": 50,
+    "confidence": 0.41523683071136475,
+    "raw_order_score": -0.87890625,
+    "confidence_bin": 3,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 0,
+    "phase": 2,
+    "provisional_token_id": 175391,
+    "default_candidate_indices": [
+      68
+    ],
+    "default_candidate_index": 68,
+    "default_sequence_position": 167,
+    "default_visual_index": 141,
+    "default_confidence": 0.9999989867210388,
+    "default_raw_order_score": 0.0,
+    "default_rank_quantile": 0.9937106370925903,
+    "confidence_gap_from_default": -0.5847621560096741,
+    "overridden_candidate_indices": [
+      19
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 3,
+    "column": 2,
+    "row_normalized": 0.2,
+    "column_normalized": 0.13333333333333333,
+    "center_distance": 7.106335201775948,
+    "local_neighbor_count": 8,
+    "local_revealed_neighbor_count": 3,
+    "local_revealed_fraction": 0.375,
+    "nearest_revealed_manhattan": 1
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1190/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "an armchair in the shape of an avocado",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281190,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 21.96398901939392,
+  "total_seconds": 21.995248317718506,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1190/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 137,
+        "sequence_position": 260,
+        "visual_index": 232,
+        "confidence": 0.4300946593284607,
+        "raw_order_score": -0.84375,
+        "confidence_bin": 3,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 14,
+        "phase": 2,
+        "provisional_token_id": 169840,
+        "default_candidate_indices": [
+          138
+        ],
+        "default_candidate_index": 138,
+        "default_sequence_position": 261,
+        "default_visual_index": 233,
+        "default_confidence": 0.9999989867210388,
+        "default_raw_order_score": 0.0,
+        "default_rank_quantile": 0.9937106370925903,
+        "confidence_gap_from_default": -0.5699043273925781,
+        "overridden_candidate_indices": [
+          137
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 14,
+        "column": 8,
+        "row_normalized": 0.9333333333333333,
+        "column_normalized": 0.5333333333333333,
+        "center_distance": 6.519202405202649,
+        "local_neighbor_count": 8,
+        "local_revealed_neighbor_count": 1,
+        "local_revealed_fraction": 0.125,
+        "nearest_revealed_manhattan": 2
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 137,
+    "sequence_position": 260,
+    "visual_index": 232,
+    "confidence": 0.4300946593284607,
+    "raw_order_score": -0.84375,
+    "confidence_bin": 3,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 14,
+    "phase": 2,
+    "provisional_token_id": 169840,
+    "default_candidate_indices": [
+      138
+    ],
+    "default_candidate_index": 138,
+    "default_sequence_position": 261,
+    "default_visual_index": 233,
+    "default_confidence": 0.9999989867210388,
+    "default_raw_order_score": 0.0,
+    "default_rank_quantile": 0.9937106370925903,
+    "confidence_gap_from_default": -0.5699043273925781,
+    "overridden_candidate_indices": [
+      137
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 14,
+    "column": 8,
+    "row_normalized": 0.9333333333333333,
+    "column_normalized": 0.5333333333333333,
+    "center_distance": 6.519202405202649,
+    "local_neighbor_count": 8,
+    "local_revealed_neighbor_count": 1,
+    "local_revealed_fraction": 0.125,
+    "nearest_revealed_manhattan": 2
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1192/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "a red cube",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281192,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 21.265228509902954,
+  "total_seconds": 21.295968055725098,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1192/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 13,
+        "sequence_position": 84,
+        "visual_index": 62,
+        "confidence": 0.44897809624671936,
+        "raw_order_score": -0.80078125,
+        "confidence_bin": 3,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 3,
+        "phase": 2,
+        "provisional_token_id": 174720,
+        "default_candidate_indices": [
+          66
+        ],
+        "default_candidate_index": 66,
+        "default_sequence_position": 174,
+        "default_visual_index": 152,
+        "default_confidence": 0.9999989867210388,
+        "default_raw_order_score": 0.0,
+        "default_rank_quantile": 0.955974817276001,
+        "confidence_gap_from_default": -0.5510208606719971,
+        "overridden_candidate_indices": [
+          13
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 3,
+        "column": 14,
+        "row_normalized": 0.2,
+        "column_normalized": 0.9333333333333333,
+        "center_distance": 7.905694150420948,
+        "local_neighbor_count": 8,
+        "local_revealed_neighbor_count": 6,
+        "local_revealed_fraction": 0.75,
+        "nearest_revealed_manhattan": 1
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 13,
+    "sequence_position": 84,
+    "visual_index": 62,
+    "confidence": 0.44897809624671936,
+    "raw_order_score": -0.80078125,
+    "confidence_bin": 3,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 3,
+    "phase": 2,
+    "provisional_token_id": 174720,
+    "default_candidate_indices": [
+      66
+    ],
+    "default_candidate_index": 66,
+    "default_sequence_position": 174,
+    "default_visual_index": 152,
+    "default_confidence": 0.9999989867210388,
+    "default_raw_order_score": 0.0,
+    "default_rank_quantile": 0.955974817276001,
+    "confidence_gap_from_default": -0.5510208606719971,
+    "overridden_candidate_indices": [
+      13
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 3,
+    "column": 14,
+    "row_normalized": 0.2,
+    "column_normalized": 0.9333333333333333,
+    "center_distance": 7.905694150420948,
+    "local_neighbor_count": 8,
+    "local_revealed_neighbor_count": 6,
+    "local_revealed_fraction": 0.75,
+    "nearest_revealed_manhattan": 1
+  }
+}

omni_diffusion/records/confirmation512/quarantine_missing_canvas/step96_q0.90/prompt_1199/omni_t2i_progressive_confidence.json

@@ -0,0 +1,119 @@
+{
+  "model_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/base_matched_pipeline_sweep/matched_training_v2/confidence_matched/checkpoint-1000",
+  "image_tokenizer_path": "/home/nitanda_sub/mark/repos/DPRM/models/showlab/magvitv2",
+  "prompt": "a green pepper",
+  "steps": 260,
+  "max_tokens": 260,
+  "alg": "entropy-penalty",
+  "temperature": 0.0,
+  "order_policy": "progressive_confidence",
+  "dprm_warmup_steps": null,
+  "dprm_table": null,
+  "dprm_action_value_model": null,
+  "dprm_order_scorer": null,
+  "dprm_action_steps": [
+    96,
+    128,
+    160,
+    192,
+    224
+  ],
+  "deployed_order_controller": null,
+  "seed": 20281199,
+  "continuation_seed": null,
+  "continuation_seed_step": 64,
+  "generation_seconds": 21.42116689682007,
+  "total_seconds": 21.46021795272827,
+  "num_image_tokens": 256,
+  "fixed_t2i_scaffold": true,
+  "text_output": "<|im_end|>\n",
+  "image_path": "/home/nitanda_sub/mark/training_runs/TPAMI-multimodal-order/omni_tpami_release_20260821/step1000_online_action_confirmation512_v1/branches/step96_q0.90/prompt_1199/omni_t2i_progressive_confidence.png",
+  "history_frame_paths": [],
+  "shared_action_canvas_path": null,
+  "counterfactual_override": {
+    "applied": true,
+    "requested_actions": 1,
+    "actions": [
+      {
+        "applied": true,
+        "step": 96,
+        "requested_quantile": 0.9,
+        "requested_rank_bin": null,
+        "candidate_count": 160,
+        "candidate_index": 66,
+        "sequence_position": 182,
+        "visual_index": 160,
+        "confidence": 0.5067753195762634,
+        "raw_order_score": -0.6796875,
+        "confidence_bin": 4,
+        "rank_quantile": 0.8993710279464722,
+        "rank_bin": 57,
+        "aux_bin": 8,
+        "phase": 2,
+        "provisional_token_id": 173629,
+        "default_candidate_indices": [
+          43
+        ],
+        "default_candidate_index": 43,
+        "default_sequence_position": 155,
+        "default_visual_index": 133,
+        "default_confidence": 0.9999989867210388,
+        "default_raw_order_score": 0.0,
+        "default_rank_quantile": 0.9622641205787659,
+        "confidence_gap_from_default": -0.4932236671447754,
+        "overridden_candidate_indices": [
+          66
+        ],
+        "shared_provisional_visual_token_ids": null,
+        "row": 10,
+        "column": 0,
+        "row_normalized": 0.6666666666666666,
+        "column_normalized": 0.0,
+        "center_distance": 7.905694150420948,
+        "local_neighbor_count": 5,
+        "local_revealed_neighbor_count": 0,
+        "local_revealed_fraction": 0.0,
+        "nearest_revealed_manhattan": 3
+      }
+    ],
+    "applied_actions": 1,
+    "step": 96,
+    "requested_quantile": 0.9,
+    "requested_rank_bin": null,
+    "candidate_count": 160,
+    "candidate_index": 66,
+    "sequence_position": 182,
+    "visual_index": 160,
+    "confidence": 0.5067753195762634,
+    "raw_order_score": -0.6796875,
+    "confidence_bin": 4,
+    "rank_quantile": 0.8993710279464722,
+    "rank_bin": 57,
+    "aux_bin": 8,
+    "phase": 2,
+    "provisional_token_id": 173629,
+    "default_candidate_indices": [
+      43
+    ],
+    "default_candidate_index": 43,
+    "default_sequence_position": 155,
+    "default_visual_index": 133,
+    "default_confidence": 0.9999989867210388,
+    "default_raw_order_score": 0.0,
+    "default_rank_quantile": 0.9622641205787659,
+    "confidence_gap_from_default": -0.4932236671447754,
+    "overridden_candidate_indices": [
+      66
+    ],
+    "shared_provisional_visual_token_ids": null,
+    "row": 10,
+    "column": 0,
+    "row_normalized": 0.6666666666666666,
+    "column_normalized": 0.0,
+    "center_distance": 7.905694150420948,
+    "local_neighbor_count": 5,
+    "local_revealed_neighbor_count": 0,
+    "local_revealed_fraction": 0.0,
+    "nearest_revealed_manhattan": 3
+  }
+}