Domains: Science
Skills: MultimodalReasoningAgenticTool Use

Four-Factor Integrity Breakdown & Primary-Source Evidence

Privacy

Weight 40%
0/10
Access Model: Public Plaintext
"We use the same 161 tasks as Handshake AI's upstream VIALS implementation, covering 174 source images and 8 artifact domains. Each model makes 3 independent repeats per task."

Originality

Weight 25%
10/10
Reuse Profile: All Original (0% reuse)
"We introduce VIALS, a visual question-answering benchmark with 161 such interpretation tasks, spanning the types of artifacts examined throughout experimental workflows in the biotech industry (rather than polished figures from publications and textbooks)."
Primary source (Paper) ↗ Verified 2026-11-17

Freshness

Weight 20%
7/10
Initial Release: 2026-08Latest Refresh: 2026-10Rolling Updates: No
"Models must be rerun on these tasks. Two prompts were updated to reduce remaining ambiguity, one reference answer and one image were updated. Released 2026-10-30 as the `v2.0.0` tag."
Primary source (HuggingFace) ↗ Verified 2026-11-17

Grading

Weight 15%
10/10
Integrity: Independent Offline (Leak-Verified)Sandbox: offline isolated
"Models receive the question and image file paths in the prompt, with the original image files available in a sandbox for code execution with Python and no network access."

Frontier Model Standings on VIALS-AA

Top 15 active models tracked
Rank Model Official Score Percentile
#1Claude Fable 5.2 max Anthropic67.1%100.0p
#2GPT-6.2 Astra max OpenAI65.8%95.5p
#3Gemini 4 Carbon high Google64.0%90.9p
#4GPT-6 Astra max OpenAI62.7%84.1p
#4GPT-6.2 Sol max OpenAI62.7%84.1p
#6Claude Opus 5.5 max Anthropic61.5%77.3p
#7GPT-6.1 Terra max OpenAI60.2%72.7p
#8Claude Sonnet 5.5 max Anthropic59.6%65.9p
#8Gemini 4 Argon high Google59.6%65.9p
#10GPT-6.1 Sol max OpenAI59.0%59.1p
#11Muse Wonder max Meta55.3%54.5p
#12Claude Haiku 5.5 max Anthropic49.1%50.0p
#13Kimi K3.1 Pro max Kimi45.3%45.5p
#14MiMo-V2.6-Pro default Xiaomi43.5%40.9p
#15Inkling 2 Small default Thinking Machines42.9%36.4p

Embeddable Audit Badge

View Markdown (.md)

Benchmark maintainers can embed this live SVG badge in GitHub READMEs or documentation:

BenchAudit VIALS-AA Tier B 5.4/10
[![BenchAudit Tier B](https://benchaudit.org/badges/vials-aa.svg)](https://benchaudit.org/benchmarks/vials-aa/)