DV-World - DV-Inter Interaction Success: leaderboard

Interaction success on the ambiguous chart requests of DV-World (2026), run in the DV-World-Agent framework: credit for clarifying turns that resolve the hidden intent of a simulated user, minus inappropriate refusals. Percent; higher is better.

Metric: Interaction success rate (%): 0.5 + 0.5 x (resolved turns minus inappropriate refusals) / (requests + 1), on ambiguous visualization requests with a GPT-5-mini user simulator, run in the DV-World-Agent framework (data manipulation, multimodal perception and proactive interaction tools); expert rubrics applied by a Gemini-2.5-Flash judge where rubric-scored, means over evaluation trials; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Grok 479.57#319
2Qwen 3 235B A22B74.62#628
3DeepSeek V3.274.05#419
4Gemini 2.5 Pro69.42#276
5GPT-5.269.25#185
6Gemini 3 Pro (Preview)66.83#119
7GPT-4.164.58#505
8Kimi K2 (Thinking)63.67
9Qwen 3 8B62.39#1873
10GLM-4.762.33#380

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=dv-world-dv-inter-interaction-success · How It Works · Data refreshed daily, snapshot 2026-11-21.