{"version":1,"type":"story","url":"https://digestai.news/story/university-of-maryland-and-aws-evaluate-gpt-6-astra-for-3d-scene-codin","json":"https://digestai.news/story/university-of-maryland-and-aws-evaluate-gpt-6-astra-for-3d-scene-codin.json","markdown":"https://digestai.news/story/university-of-maryland-and-aws-evaluate-gpt-6-astra-for-3d-scene-codin.md","slug":"university-of-maryland-and-aws-evaluate-gpt-6-astra-for-3d-scene-codin","headline":"University of Maryland and AWS evaluate GPT-6 Astra for 3D scene coding, hit 53% indoor accuracy","summary":"The agent writes, runs, and revises code iteratively, producing a program that captures objects, geometry, layout, and camera position. To measure performance they built LEGO‑Bench, a benchmark of 208 images from 104 indoor and outdoor scenes rendered from simulator ground truth.\n\nSix GPT configurations were tested. All delivered a usable scene, but geometric accuracy varied widely. GPT‑6 Astra, the strongest model, achieved 53.4 % accuracy on indoor scenes and 39.6 % on outdoor scenes, while weaker setups hovered around 15 %. Raising the reasoning budget boosted Astra’s office subset score from 32.3 % to 61.8 %. The authors found agents struggled with self‑assessment, often mis‑ranking revisions. Their LEGO‑Plugin, which replaces self‑judgment with concrete measurements, lifted weaker models by up to 62.7 % and gave the top model a modest two‑point gain.\n\nThe generated scenes can be queried for vision tasks, yielding roughly half the object‑detection performance of the specialized DINO model, but larger gaps for segmentation and depth compared with SAM 3 and Depth Anything 3. Researchers see GPT‑6 Astra’s spatial understanding as a notable step, while industry players like Unity are already releasing plugins for agents such as Claude Code and Codex.","keyPoints":["LEGO‑Bench benchmark uses 208 rendered images with hidden ground‑truth geometry for automated scoring.","GPT‑6 Astra scored 53.4% indoor and 39.6% outdoor accuracy; weaker GPT configs stayed near 15%.","LEGO‑Plugin replaces self‑assessment with concrete measurements, improving weaker agents by up to 62.7%."],"whyItMatters":"Accurate 3D reconstruction from photos enables new workflows for designers, game developers, and robotics, while highlighting current limits of AI self‑evaluation and the need for measurement‑driven refinement.","category":{"slug":"agents","name":"Agents & Tools","url":"https://digestai.news/category/agents"},"entities":{"companies":["University of Maryland","Unity","World Labs","Google DeepMind"],"models":["GPT-6 Astra","Claude Code","Codex","Atlas","GenCeption"],"people":["Yoav Artzi"]},"firstPublishedAt":"2026-10-03T08:31:04Z","updatedAt":"2026-10-03T08:31:04Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"The Decoder","title":"AI agents build 3D scenes from photos but have no idea if they got it right","url":"https://the-decoder.com/ai-agents-build-3d-scenes-from-photos-but-have-no-idea-if-they-got-it-right","publishedAt":"2026-10-03T08:31:04Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"AI Drives Cars and Maps Interiors","url":"https://digestai.news/thread/openais-gpt-6-astra-drives-toyota-corolla-through-parking-lot-course","storyCount":2},"cite":{"text":"Digest AI, \"University of Maryland and AWS evaluate GPT-6 Astra for 3D scene coding, hit 53% indoor accuracy\", 3 October 2026, https://digestai.news/story/university-of-maryland-and-aws-evaluate-gpt-6-astra-for-3d-scene-codin","publisher":"Digest AI","title":"University of Maryland and AWS evaluate GPT-6 Astra for 3D scene coding, hit 53% indoor accuracy","datePublished":"2026-10-03T08:31:04Z","url":"https://digestai.news/story/university-of-maryland-and-aws-evaluate-gpt-6-astra-for-3d-scene-codin"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}