SHROOM-Visions 2026
Models inspect an image and a written response. They mark unsupported characters, estimate the probability of an error and classify each error in four languages.
The organizers report 27 teams and 623 submissions. SKstars ran two Qwen2.5-VL models. The final tables score span overlap and character-level correlation.
- Made by
- SHROOM-Visions organizers
- Released
- April 2026
- Kind
- Faithful to sources
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Prediction
- Tasks
- Public, plus a private test set
- Measures
- Capability
- Interaction
- Single turn
- Items from
- Generated
- Refresh
- One fixed set
- Run by
- A host
- In views
- Vision
Gemma 4 + Mistral-Small + Qwen3-VL + Qwen3.6 tops the leaderboard with 0.46 No scores yet
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The organizers report no overall score, so this is the plain mean of labeled correlation across four languages. Where a model ran on its own with several settings, we keep its best. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. Agents built around a model, such as TÜRKSAT + Gemini 3.5 Flash, are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best.
- 0.46 Gemma 4
- 0.45 TÜRKSAT + Gemini 3.5 Flash
- 0.42 champ + Qwen3.5-27B
- 0.42 dynamos (Qwen2.5-VL-7B + Qwen2.5-VL-32B)
- 0.40 SmurfCat + Qwen3-VL-8B-Instruct
- 0.39 eye-be-am (Qwen3.5-4B + Qwen3.5 397B-A17B)
- 0.38 caml + Qwen3.5-4B
- 0.36 HalluVision + Qwen2.5-VL-3B
- 0.36 Bit-by-bit (Qwen2.5-VL-7B + InternVL3.5-8B)
- 0.34 Sckwoky + Gemma 4 12B
- 0.32 XLM-RoBERTa-large
- 0.29 SpanCalib-VLM
- 0.27 MycoGate-VL
- 0.27 CPS Lab + Qwen3-VL-8B-Instruct
- 0.25 mDeBERTa
- 0.24 XLM-RoBERTa
- 0.19 WINNAR749 + Qwen2.5-VL-3B
- 1 Gemma 4 + Mistral-Small + Qwen3-VL + Qwen3.6 0.46
- 2 TÜRKSAT + Gemini 3.5 Flash 0.45
- 3 champ + Qwen3.5-27B 0.42
- 4 dynamos (Qwen2.5-VL-7B + Qwen2.5-VL-32B) 0.42
- 5 SmurfCat + Qwen3-VL-8B-Instruct 0.40
- 6 eye-be-am (Qwen3.5-4B + Qwen3.5 397B-A17B) 0.39
- 7 caml + Qwen3.5-4B 0.38
- 8 HalluVision + Qwen2.5-VL-3B 0.36
- 9 Bit-by-bit (Qwen2.5-VL-7B + InternVL3.5-8B) 0.36
- 10 Sckwoky + Gemma 4 12B 0.34
- 11 XLM-RoBERTa-large + Qwen2-VL-2B-Instruct 0.32
- 12 SpanCalib-VLM 0.29
- 13 MycoGate-VL 0.27
- 14 CPS Lab + Qwen3-VL-8B-Instruct 0.27
- 15 mDeBERTa + SigLIP 0.25
- 16 XLM-RoBERTa + SigLIP 0.24
- 17 WINNAR749 + Qwen2.5-VL-3B 0.19
Source: Official SHROOM-Visions 2026 leaderboard
| # | Entry | Score | Setting | Submitted by | Date | Source |
|---|---|---|---|---|---|---|
| 1 | Gemma 4 + Mistral-Small + Qwen3-VL + Qwen3.6 | 0.46 | vroom-vroom | vroom-vroom | Official leaderboard | |
| 2 | TÜRKSAT + Gemini 3.5 Flash | 0.45 | TÜRKSAT | May 2026 | Official leaderboard | |
| 3 | champ + Qwen3.5-27B | 0.42 | champ | Feb 2026 | Official leaderboard | |
| 4 | dynamos (Qwen2.5-VL-7B + Qwen2.5-VL-32B) | 0.42 | dynamos | Jan 2025 | Official leaderboard | |
| 5 | SmurfCat + Qwen3-VL-8B-Instruct | 0.40 | SmurfCat | Oct 2025 | Official leaderboard | |
| 6 | eye-be-am (Qwen3.5-4B + Qwen3.5 397B-A17B) | 0.39 | eye-be-am | Feb 2026 | Official leaderboard | |
| 7 | caml + Qwen3.5-4B | 0.38 | caml | Feb 2026 | Official leaderboard | |
| 8 | HalluVision + Qwen2.5-VL-3B | 0.36 | HalluVision | Jan 2025 | Official leaderboard | |
| 9 | Bit-by-bit (Qwen2.5-VL-7B + InternVL3.5-8B) | 0.36 | Bit-by-bit | Jan 2025 | Official leaderboard | |
| 10 | Sckwoky + Gemma 4 12B | 0.34 | Sckwoky | Jun 2026 | Official leaderboard | |
| 11 | XLM-RoBERTa-large + Qwen2-VL-2B-Instruct | 0.32 | USP | USP | Official leaderboard | |
| 12 | SpanCalib-VLM | 0.29 | SIT | SIT | Official leaderboard | |
| 13 | MycoGate-VL | 0.27 | L1cache | L1cache | Official leaderboard | |
| 14 | CPS Lab + Qwen3-VL-8B-Instruct | 0.27 | CPS Lab | Oct 2025 | Official leaderboard | |
| 15 | mDeBERTa + SigLIP | 0.25 | Scalar_Nitk | Scalar_Nitk | Official leaderboard | |
| 16 | XLM-RoBERTa + SigLIP | 0.24 | DSR | DSR | Official leaderboard | |
| 17 | WINNAR749 + Qwen2.5-VL-3B | 0.19 | WINNAR749 | Jan 2025 | Official leaderboard |
| # | Model | Score | Setting | Submitted by | Date | Source |
|---|
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Factuality
- SimpleVQA In use Models answer 2,025 short factual questions about images across 9 tasks and 9 topics. An LLM judge grades answers against fixed references.
Qwen3.7-Plus 81.7%
- HallusionBench In use Models answer 1,129 yes or no questions about 346 images, built in pairs to expose language hallucination and visual illusion. Scored by accuracy.
Keye-VL-2.0-30B-A3B 73.5%
- SimpleQA Verified In use 1,000 human-verified, de-duplicated short-answer factual questions measuring parametric knowledge.
Gemini 3.1 Pro 77.5%
- AA-Omniscience v1.1 In use Models answer 3,000 knowledge questions in six domains without tools and may abstain. The index adds correct answers and subtracts wrong ones.
Inkling 2 Small 16.1
- GaRAGe In use Models write long-form RAG answers to 2,366 questions with human-annotated grounding passages. Scoring checks whether answers use the relevant passages.
Amazon Nova Pro 60.7%
- Kimi Hallu Bench In use Models answer hard prompts drawn from practical work tasks. Fact-checking agents score the share of claims that are hallucinated. GPT-6 Astra 4.20%
Data audit
All 8 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access Public tasks, plus a private test set.
- Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
- Pass Independent runs Scores from the official leaderboard (17 entries).
- Pass Models scored 17 entries on its leaderboard, none a model we track.
- Note Newest score No scores yet.
- Note Frontier models No models scored yet.
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on helsinki-nlp.github.io: 11 rows for models we track and 6 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.