HiEviDR-Bench - Wiki (Deep Research): leaderboard
Metric: Overall score (0-100): sum of five 0-20 dimension scores (multimodal report quality, evidence traceability, and progressively gated citation, claim and answer scores) judged by Qwen3-VL-235B-A22B-Instruct against the hierarchical evidence graph, on the 1,000-question Wikipedia subset; MLLM with Deep Research: up to three rounds of question refinement with top-20 retrieval, re-ranked to 25 text chunks and 5 images; higher is better. Source: arxiv.org. Saturation forecast: Around 2034. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Grok 4.5 | 46.41 |
| 2 | GPT-5.6 Sol | 45.13 |
| 3 | Gemma 4 31B (IT) | 43.71 |
| 4 | GPT-5 | 43.31 |
| 5 | Qwen 3.5 27B | 42.44 |
| 6 | GPT-5 Mini | 42.25 |
| 7 | Qwen 3.5 35B A3B | 41.94 |
| 8 | Qwen 3.5 4B | 40.35 |
| 9 | gemma-4-E4B-it | 40.25 |
| 10 | Qwen 3.5 9B | 39.88 |
| 11 | Qwen 3 VL 8B Instruct | 39.12 |
| 12 | InternVL3.5-8B | 38.07 |
Interactive version: theaggregate.ai/benchmark?slug=hievidr-bench-wiki-deep-research · How It Works · Data refreshed daily, snapshot 2026-09-29.