Add vultr/VultronRetrieverFlash-Qwen3.5-0.8B ViDoRe results

[athrael-soju]

ViDoRe V1/V2/V3 task results for vultr/VultronRetrieverFlash-Qwen3.5-0.8B, the 0.8B small-tier late-interaction retriever (sibling of vultr/VultronRetrieverPrime-Qwen3.5-8B).

Depends on embeddings-benchmark/mteb#4845 (the ModelMeta) — please merge that first so the model name resolves via get_model_meta.

• 22 task JSONs + model_meta.json under results/vultr__VultronRetrieverFlash-Qwen3.5-0.8B/5d1a696e8e62f12508045a93543dfd0488ea3b77/.
• Produced with the MTEB late-interaction evaluator (mteb 2.12.30), dim 320 / 1792 visual tokens; the JSONs are taken from the model repo eval_results/ directory.
• Coverage: ViDoRe V1 (10, ndcg@5), V2 (4, ndcg@5), V3 (8/10 domains, ndcg@10 — Telecom + Nuclear not evaluated, same as the 8B Prime entry). No .v2 task names.

Means over the submitted JSONs: V1 0.8815 / V2 0.6036 / V3 0.5649.

[athrael-soju]

🕺

[github-actions]

Model Results Comparison

Reference models: intfloat/multilingual-e5-large, google/gemini-embedding-001
New models evaluated: vultr/VultronRetrieverFlash-Qwen3.5-0.8B

Results for vultr/VultronRetrieverFlash-Qwen3.5-0.8B

task_name	vultr/VultronRetrieverFlash-Qwen3.5-0.8B	Max result	Model with max result	In Training Data
Vidore2BioMedicalLecturesRetrieval	.598	.670	DataScience-UIBK/Argus-Colqwen3.5-9b-v0-bf16	False
Vidore2ESGReportsHLRetrieval	.664	.791	DataScience-UIBK/Argus-Colqwen3.5-9b-v0-bf16	False
Vidore2ESGReportsRetrieval	.574	.660	OpenSearch-AI/Ops-Colqwen3-4B	False
Vidore2EconomicsReportsRetrieval	.579	.658	DataScience-UIBK/Argus-Colqwen3.5-9b-v0	False
Vidore3ComputerScienceRetrieval	.738	.809	webAI-Official/webAI-ColVec1-9b	False
Vidore3EnergyRetrieval	.611	.703	vultr/VultronRetrieverPrime-Qwen3.5-8B	False
Vidore3FinanceEnRetrieval	.606	.690	vultr/VultronRetrieverPrime-Qwen3.5-8B	False
Vidore3FinanceFrRetrieval	.408	.545	vultr/VultronRetrieverPrime-Qwen3.5-8B	False
Vidore3HrRetrieval	.584	.700	webAI-Official/webAI-ColVec1-9b	False
Vidore3IndustrialRetrieval	.462	.574	vultr/VultronRetrieverPrime-Qwen3.5-8B	False
Vidore3PharmaceuticalsRetrieval	.631	.682	vultr/VultronRetrieverPrime-Qwen3.5-8B	False
Vidore3PhysicsRetrieval	.479	.517	vultr/VultronRetrieverPrime-Qwen3.5-8B	False
VidoreArxivQARetrieval	.886	.938	VAGOsolutions/SauerkrautLM-ColQwen3-8b-v0.1	True
VidoreDocVQARetrieval	.622	.687	webAI-Official/webAI-ColVec1-9b	True
VidoreInfoVQARetrieval	.922	.952	webAI-Official/webAI-ColVec1-9b	True
VidoreShiftProjectRetrieval	.804	.947	DataScience-UIBK/Argus-Colqwen3.5-4b-v0	False
VidoreSyntheticDocQAAIRetrieval	.977	1.000	athrael-soju/colqwen3.5-4.5B-v3	True
VidoreSyntheticDocQAEnergyRetrieval	.967	.980	nvidia/llama-nemotron-colembed-vl-3b-v2	True
VidoreSyntheticDocQAGovernmentReportsRetrieval	.962	.989	DataScience-UIBK/Argus-Colqwen3.5-9b-v0-bf16	True
VidoreSyntheticDocQAHealthcareIndustryRetrieval	.980	1.000	VAGOsolutions/SauerkrautLM-ColQwen3-4b-v0.1	True
VidoreTabfquadRetrieval	.918	.981	nvidia/nemotron-colembed-vl-4b-v2	True
VidoreTatdqaRetrieval	.779	.857	DataScience-UIBK/Argus-Colqwen3.5-9b-v0-bf16	True
Average	.716	.788	nan	-

---