{"submission_id":"3af96b5e-3b77-44d4-8a04-87f53fc0f162","submission_revision_id":"b8a6a21f-53da-462e-bd06-5af9b5259fd7","revision_version":8,"submission_status":"under_review","submitted_at":"2026-09-13T14:38:30.351694Z","submitter_name":"Yichen Wu","submitter_email":"myrealnameisk@outlook.com","task":{"status":"submitted","version":8,"extra_fields":{"difficulty_self_test":{"model":"GPT-6 Astra","harness":"Codex","estimated_score":"<1%"}},"submitted_at":"2026-09-13T14:38:30.351694+00:00","industry_domain":"Presentation Design, Data Visualization, Human-AI Systems","task_short_name":"editable_experience_evolution_infographic_alignment_1","operating_system":"Windows","task_description":"Objective: Recreate the infographic in `input.png` as a single-slide PowerPoint with identical visual appearance and independently editable layout components.\n\nStarting State: A canonical 1200x896 RGB full-slide reference image (`input.png`) and five supplied chart images: `reconstructed.png`, `pillow_reconstructed.png`, `reconstructed_pillow.png`, `pillow_hexagon.png`, and `hexagon_heatmap_vector.png`.\n\nInstructions & Rules: Use `input.png` only as the visual reference and do not insert it into the output slide. Recreate the complete layout, spacing, alignment, panels, arrows, labels, text boxes, and visible text.\n\nNative-editable components: all ordinary text outside the supplied chart images, text boxes, panel backgrounds, speech-bubble frames, arrows, dividers, labels, chart containers, and other ordinary PowerPoint graphic elements must be independently editable PowerPoint objects.\n\nPermitted chart assets: the five supplied chart PNGs may be inserted as separate image objects in their corresponding chart regions. All pixels inside those five chart images are covered by this exception, including chart geometry, chart labels, legends, axis labels, and annotations. The chart image objects themselves must remain independently movable and resizable.\n\nIn particular, `reconstructed_pillow.png` may be inserted directly as the line-chart region. Its agent labels around the line chart, including labels such as `Agent 1`, `Agent 2`, and `Agent N`, and its right-side legend are intentionally part of the Python/Pillow-generated image. They do not need to be recreated as separate editable PowerPoint text boxes and are evaluated as part of the permitted chart image.\n\nPermitted cropped assets: the character illustrations, brain illustrations, the isolated document-and-pencil pictogram near the Baseline Assessment panel, and only the pictograms inside the five icons under “Continuous Experience Injection” — Study, Socializing, Daily Life, Milestones, and Challenges — may be cropped directly from `input.png` and remain non-editable image objects.\n\nOnly the pictograms may be cropped from the five Continuous Experience Injection icons. Their speech-bubble frames and all associated text must remain editable PowerPoint objects. The document-and-pencil pictogram must be cropped as an isolated icon without its surrounding panel, chart, text, or other slide content.\n\nNo separate source files are required for the permitted cropped assets. Do not insert `input.png`, the complete reference image, or any other full-slide or large-region screenshot. Do not use any raster image that is not one of the five supplied chart images or one of the explicitly permitted cropped assets.","software_licensing":"Commercial","evaluation_criteria":null,"verification_method":"Use the solver-visible `input.png` as the canonical visual target. Use the hidden `recreated_infographic.pptx` only for slide-structure, asset-policy, and editability verification. The hidden reference PPTX must not be available to the solver.\n\nThe canonical `input.png` was generated from the reference PPTX using Microsoft PowerPoint COM automation. The submitted PPTX must be rendered using the same PowerPoint version, fonts, scaling, and complete-slide export procedure:\n\n$powerPoint = New-Object -ComObject PowerPoint.Application\n$powerPoint.Visible = $false\n$presentation = $powerPoint.Presentations.Open($pptxPath, -1, -1, 0)\n\nif ($presentation.Slides.Count -ne 1) {\n    throw \"The submitted presentation must contain exactly one slide.\"\n}\n\n$slide = $presentation.Slides.Item(1)\n$slide.Export($outputPng, \"PNG\", 1200, 896)\n\n$presentation.Close()\n$powerPoint.Quit()\n\nThe export must be the complete slide canvas at exactly 1200x896 pixels. Do not use screen capture, selected-object export, object-bound cropping, stretching, resizing, or PDF rasterization for the primary visual comparison.\n\nThe source slide must have a 900x672 point canvas, or an exactly equivalent 1200:896 aspect ratio. The rendered candidate must be decoded as a uint8 RGB array with shape 896x1200x3. No resizing, resampling, alpha compositing, or color conversion may be applied before comparison.\n\nLet `A` be the canonical `input.png` array and `B` be the rendered candidate array.\n\nExact pixel match rate is:\n\n`number of pixels for which all three RGB channels in A and B are identical / (1200*896)`\n\nMean absolute error is:\n\n`mean(abs(A-B))` over all RGB channels.\n\nRMSE is:\n\n`sqrt(mean((A-B)^2))` over all RGB channels.\n\nSSIM must be computed with:\n\n`skimage.metrics.structural_similarity(A, B, channel_axis=2, data_range=255, win_size=11, gaussian_weights=True, sigma=1.5, use_sample_covariance=False)`\n\nThe following visual thresholds are mandatory:\n\n- output dimensions = exactly 1200x896 RGB;\n- exact pixel match rate >= 90%;\n- SSIM >= 0.98;\n- mean absolute error <= 5.0 on the 0–255 RGB scale;\n- RMSE <= 15.0 on the 0–255 RGB scale;\n- global foreground-mask IoU >= 0.98;\n- each predefined major-region IoU >= 0.95;\n- every fixed key edge and anchor center must be within 3 pixels of the canonical target.\n\nMaximum per-channel absolute difference, exact pixel match rate, MAE, RMSE, SSIM, global foreground IoU, and per-region IoU must all be reported. Maximum per-channel absolute difference is diagnostic and is not required to be zero.\n\nThe following coordinate manifest is binding:\n\n{\n  \"canvas\": [0, 0, 1200, 896],\n  \"stage_header_a\": [0, 0, 389, 90],\n  \"stage_header_b\": [390, 0, 807, 90],\n  \"stage_header_c\": [808, 0, 1200, 90],\n  \"left_top_panel\": [8, 102, 376, 388],\n  \"left_bottom_panel\": [10, 469, 378, 882],\n  \"center_injection_header\": [403, 102, 797, 154],\n  \"center_icon_row\": [405, 173, 796, 290],\n  \"center_line_chart_region\": [403, 171, 797, 544],\n  \"center_state_evolution_panel\": [403, 556, 797, 888],\n  \"right_assessment_panel\": [824, 102, 1192, 348],\n  \"right_analysis_panel\": [824, 379, 1192, 888],\n  \"left_baseline_arrow\": [181, 394, 204, 458],\n  \"right_assessment_arrow\": [996, 348, 1020, 380]\n}\n\nCoordinate convention:\n\n- The origin `(0,0)` is the top-left pixel.\n- `x` increases from left to right.\n- `y` increases from top to bottom.\n- Rectangles use half-open coordinates `[x0, y0, x1, y1)`.\n- The left and top boundaries are included.\n- The right and bottom boundaries are excluded.\n- Anchor coordinates refer to pixel-center positions.\n- The evaluator must use these rectangles exactly and must not redefine, move, resize, or optimize them.\n\nEvery edge of every rectangle in the manifest is a fixed key edge.\n\nThe fixed key anchor centers are:\n\n{\n  \"left_character\": [188, 278],\n  \"left_baseline_radar\": [184, 684],\n  \"left_top_brain\": [308, 583],\n  \"left_bottom_brain\": [61, 724],\n  \"left_document_pencil\": [320, 720],\n  \"study_pictogram\": [435, 214],\n  \"socializing_pictogram\": [512, 214],\n  \"daily_life_pictogram\": [589, 214],\n  \"milestones_pictogram\": [665, 214],\n  \"challenges_pictogram\": [751, 214],\n  \"line_chart_node_1\": [438, 488],\n  \"line_chart_node_2\": [509, 470],\n  \"line_chart_node_3\": [577, 449],\n  \"line_chart_node_4\": [643, 421],\n  \"line_chart_node_5\": [710, 381],\n  \"line_chart_node_6\": [778, 327],\n  \"center_brain\": [535, 716],\n  \"center_feedback_loop\": [680, 722],\n  \"right_character\": [900, 255],\n  \"right_radar\": [1070, 239],\n  \"right_scatterplot\": [922, 718],\n  \"right_heatmap\": [1088, 714]\n}\n\nFor each fixed region, create foreground masks using the rule that a pixel is foreground when at least one RGB channel is below 250. Compute:\n\n`IoU = intersection(mask_reference, mask_candidate) / union(mask_reference, mask_candidate)`\n\nAll visible text must match the canonical target in spelling, capitalization, punctuation, line breaks, and reading order.\n\nText embedded inside the five permitted chart PNGs is evaluated as part of those chart images and is exempt from native PowerPoint text-box editability. In particular, the agent labels and right-side legend embedded inside `reconstructed_pillow.png` may remain inside that image.\n\nEvery native-editable component must be independently selectable, editable, movable, and deletable. Editing one component must not modify unrelated components.\n\nThe five supplied chart PNGs, the character illustrations, the brain illustrations, the isolated document-and-pencil pictogram, and the five specified Continuous Experience Injection pictograms must remain separate image objects if used. The image objects must be independently movable and resizable.\n\nPermitted cropped assets may be extracted directly from `input.png`. The five Continuous Experience Injection speech-bubble frames and all associated text must remain native-editable PowerPoint objects.\n\nThe full `input.png`, any full-slide screenshot, any large-region screenshot, and any unlisted raster image are prohibited.\n\nThe submitted PPTX must contain no clipping, unintended overlap, misplaced connectors, missing components, or material visual differences from `input.png`.\n\nThe actual editable PPTX must be submitted and inspected. A rendered PNG or PDF alone is insufficient evidence.","software_and_version":"Microsoft PowerPoint 365 (desktop), Python 3.11, python-pptx 1.x","specific_requirements":null,"expected_completion_time":null},"files":[{"file_id":"55001212-29fd-48f8-af11-388a4f398143","filename":"reconstructed.png","category":"input_materials","content_type":"image/png","size_bytes":13018},{"file_id":"d1fcba03-a983-4eff-bf36-4ddb4ff41a49","filename":"hexagon_heatmap_vector.png","category":"input_materials","content_type":"image/png","size_bytes":12872},{"file_id":"55635e90-2a96-4740-b468-12bccb1a58d4","filename":"pillow_reconstructed.png","category":"input_materials","content_type":"image/png","size_bytes":2443},{"file_id":"9344bd05-4e41-4b4c-86b5-8b03825e12b4","filename":"pillow_hexagon.png","category":"input_materials","content_type":"image/png","size_bytes":20279},{"file_id":"84d0ca01-49c6-4c5e-9099-593336526a82","filename":"reconstructed_pillow.png","category":"input_materials","content_type":"image/png","size_bytes":9858},{"file_id":"1fbe5dff-ec6e-49d0-8d3d-ad0f92fd5b06","filename":"recreated_infographic.pptx","category":"reference_output","content_type":"application/vnd.openxmlformats-officedocument.presentationml.presentation","size_bytes":159230},{"file_id":"436cf42d-7c23-4de3-bbd7-c328394579cc","filename":"input.png","category":"input_materials","content_type":"image/png","size_bytes":405133},{"file_id":"00859c86-314f-493d-b552-373acd89e09e","filename":"input.png","category":"reference_output","content_type":"image/png","size_bytes":405133},{"file_id":"33a08cca-8185-4df6-ae4b-a68ad2520e8d","filename":"recreated_infographic_final_v2.png","category":"difficulty_self_test_evidence","content_type":"image/png","size_bytes":286755}],"deep_review":{"overall_verdict":"Strong Accept","review_text":"**Overall Verdict:** Strong Accept — The submission is complete, coherent, reachable, and objectively verifiable.\n\n**Completeness Check:**\n- Instruction: Complete — Detailed recreation objective, editability rules, permitted assets, coordinate manifest, and export/evaluation requirements are provided.\n- Inputs: Complete — All five chart PNGs, both identical `input.png` entries, and the self-test PNG are present, readable, correctly sized, and consistent with their stated roles.\n- Toolchain: Complete — PowerPoint 365, Python 3.11, `python-pptx`, and Windows provide a legally obtainable workflow.\n- Reference output: Complete — `recreated_infographic.pptx` is a valid, unencrypted one-slide OOXML package with the required 900×672-point canvas, 87 native shape elements, 17 separate image objects, and exact matches for the five supplied chart assets.\n- Rubric: Complete — Exact pixel, SSIM, MAE, RMSE, foreground-IoU, regional-IoU, edge, anchor, structure, and editability criteria are specified.\n\n**File and Task Reachability:**\nAttachment classification: `reachable`. Task classification: `reachable`. I inspected `reconstructed.png`, `hexagon_heatmap_vector.png`, `pillow_reconstructed.png`, `pillow_hexagon.png`, `reconstructed_pillow.png`, both manifest entries named `input.png`, `recreated_infographic.pptx`, and `recreated_infographic_final_v2.png`. Byte counts match the manifest; image formats and dimensions are valid; the PPTX has safe internal paths, no external targets, no macros, and no full-slide raster embedded. No manifest content was left unread. Microsoft PowerPoint COM rendering and the final pixel-metric computation were not run because that desktop tooling is unavailable in this review environment; this is a reviewer-tooling limitation, not a submission defect. The end-to-end path from supplied images to an editable PPTX and exact exported evaluation image is otherwise clear.\n\nLicense keys, accounts, tokens, and other credentials are not required from the contributor.\n\nConfidential, proprietary, or internal material may be evaluated as submitted evidence without requiring additional disclosure.\n\n**Special Screening:**\nNone.\n\n**Recommended Changes:**\nThe submission provides a strong, professionally specified task with authentic, usable evidence and a rigorous evaluator contract. No changes are required.","reviewed_at":"2026-09-13T14:46:29.078958Z"},"claim":null}