Undominated.ai

All rankings / Rerankers

Reranking · Computer science

ViDoRe 3 reranking results from MTEB: English queries, the test split, and BM25s first-stage retrieval. Each subject has its own board.

8 entries · nDCG@10 × 100 · Higher is better

Source publication date not supplied. Retrieved 2026-11-05.

Source: MTEB / embeddings-benchmark · CC0-1.0.

Point-estimate ordering. The source supplies no confidence bounds, so small score differences are not evidence of a decisive lead.

Reranking · Computer science · nDCG@10 × 100
Score rankModel / evaluated variantnDCG@10 × 100
1nvidia/Nemotron-3.5-Rerank-8B-BF16nvidiaCheckpoint 708e844c3da080.503
2Qwen/Qwen3-VL-Reranker-8BQwenCheckpoint 407993db318a79.118
3KaLM-Embedding/KaLM-Reranker-V1-Large-R2KaLM-EmbeddingCheckpoint a862dd8581d479.023
4Qwen/Qwen3-VL-Reranker-2BQwenCheckpoint 4bd860ac4f1576.921
5nvidia/llama-nemotron-rerank-vl-1b-v2nvidiaCheckpoint b8a9987b05b776.854
6KaLM-Embedding/KaLM-Reranker-V1-Small-R2KaLM-EmbeddingCheckpoint 68b74f09ff2a76.452
7KaLM-Embedding/KaLM-Reranker-V1-Nano-R2KaLM-EmbeddingCheckpoint 32b242bb7fd775.59
8BAAI/bge-reranker-v2-m3BAAICheckpoint 953dc6f6f85a68.703

What this ranking can tell you

  • These are document retrieval rerankers evaluated on a fixed candidate set, not a general language-model ranking.
  • Coverage is limited to model revisions with this exact published task and English subset. Missing results are not scored.
  • The source supplies no confidence interval here; the displayed rank orders point estimates and does not establish statistical significance.
  • No result publication date is supplied. The retrieval date records when the pinned result files were checked.

Check the source

The original evidence is retained byte for byte. The downloadable board includes the source URLs, revision and SHA-256 hashes.

Download this board and provenance →

Source files and licences

Explore other model types →Read the benchmark guide →

Evidence & Ask
Markdown for LLMs