# Elorian — Introducing ScienceVision

- Company: Elorian (elorian.ai)
- Announced: 2026-10-29
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://elorian.ai/blog/sciencevision
- Record: https://forck.live/items/31170-introducing-sciencevision
- Subject: Glimpse visual reasoning models
- Models affected: GPT-6.1 Sol, Kimi K3, GPT-6 Astra, Claude Fable 5.1

Elorian AI and UniPat AI built ScienceVision, a 250-task benchmark that asks models to find, measure and interpret structures in 3D scientific artifacts. The tasks are split fifty each across biology, chemistry, medical imaging, mechanical CAD and architecture, spanning 22 subdomains, and each keeps the artifact in its native molecular, volumetric, CAD or building representation. The benchmark compares three kinds of access to the same artifact: a coding agent with the native file, the same plus an inspection interface exposed through the Model Context Protocol, and six pre-rendered views that the solver cannot change. Source access beat fixed views for every model tested, without exception, but adding the inspection tools helped eight of twelve models and hurt four, with margins from +2.17 points for GPT-6.1 Sol to −2.40 for Kimi K3. The leader changes with the setting: GPT-6 Astra tops the source-access settings at 0.53, while Claude Fable 5.1 tops the fixed-view condition at 0.30. Biology holds the highest single domain score at 0.65, and no model falls below 0.24 in biology or chemistry, while mechanical CAD, medical imaging and architecture have the lowest floors, with their weakest models under 0.10. Scores are averaged over the available repeats within each task, then across the 250 tasks, and all 36 reported configurations carry a score for every task.

## Evidence

Verbatim from https://elorian.ai/blog/sciencevision:

> A scientific question about a three-dimensional object asks three things at once: find the structure the question is about, measure its geometry, and interpret what you measured using what you know about the field. Fixed images, the native data file, and interactive tools each give a different amount of help with the first two.
> 
> The questions are concrete. Find the pipe bundle furthest from the origin in a building model, then count how many independent pipes it contains, not how many segments they are drawn in. Locate a target in a CT volume, measure its extent, and report a signed centroid offset under a specified coordinate convention. Neither can be answered from a picture of the object.
> 
> ScienceVision asks 250 of them. Fifty each in biology, chemistry, medical imaging, mechanical CAD and architecture, across 22 subdomains. Each task keeps the artifact in its native representation (molecular, volumetric, CAD or building) and pairs it with a question and a task-specific reference answer.
> 
> “Expert-level” here means the domain knowledge and conventions needed to interpret these objects.

---

Record: https://forck.live/items/31170-introducing-sciencevision
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md