Japanese decision models saihai (采配)
Yuichi Tateno PRO
hotchpotch
AI & ML interests
Information Retrieval with LLMs
Recent Activity
updated a Space about 3 hours ago
hotchpotch/S1MB-leaderboard new activity about 15 hours ago
hotchpotch/s1mb-result:Add complete local multimodal measurements for mohit67890__imajev-9b new activity about 15 hours ago
hotchpotch/s1mb-result:Add complete local multimodal measurements for mohit67890__imajev-2bOrganizations
Compact multilingual embedding models with 8M and 25M active parameters, including pretrained checkpoints and retrieval-fine-tuned releases.
-
hotchpotch/bekko-embedding-v1-a8m
Sentence Similarity • 0.1B • Updated • 7.34k • • 23 -
hotchpotch/bekko-embedding-v1-a25m
Sentence Similarity • 0.1B • Updated • 18.1k • • 32 -
bekko-embedding-web
🚀4Run ultra-compact multilingual embeddings in your browser.
-
Bekko Embedding: Parameter-Efficient Multilingual Retrieval with Ultra-Compact Encoders
Paper • 2607.25180 • Published • 5
FineWeb2 Edu Japanese: A high-quality, filtered Japanese dataset (120M texts, 89.3B tokens) for educational AI training.
日本語SPLADEモデル
Ultra small system one models
OpenProvence is a fully open-source implementation of Provence-style reranker–pruner models. These models remove irrelevant context while simultaneous
-
Open Provence Demo
✂4Context pruning demo for context engineering and RAG
-
hotchpotch/open-provence-reranker-xsmall-v1
Text Ranking • 36.8M • Updated • 467 • 1 -
hotchpotch/open-provence-reranker-v1
Text Ranking • 0.1B • Updated • 235 -
hotchpotch/open-provence-reranker-v1-gte-modernbert-base
Text Ranking • 0.1B • Updated • 280 • 2
日本語rerankerシリーズ
-
hotchpotch/japanese-reranker-tiny-v2
Text Ranking • 29.4M • Updated • 570 • 6 -
hotchpotch/japanese-reranker-xsmall-v2
Text Ranking • 36.8M • Updated • 33.3k • • 7 -
hotchpotch/japanese-reranker-small-v2
Text Ranking • 70.2M • Updated • 23.3k • • 4 -
hotchpotch/japanese-reranker-base-v2
Text Ranking • 0.1B • Updated • 2.22k • • 8
情報検索のための質問文自動作成モデル
Japanese decision models saihai (采配)
Ultra small system one models
Compact multilingual embedding models with 8M and 25M active parameters, including pretrained checkpoints and retrieval-fine-tuned releases.
-
hotchpotch/bekko-embedding-v1-a8m
Sentence Similarity • 0.1B • Updated • 7.34k • • 23 -
hotchpotch/bekko-embedding-v1-a25m
Sentence Similarity • 0.1B • Updated • 18.1k • • 32 -
bekko-embedding-web
🚀4Run ultra-compact multilingual embeddings in your browser.
-
Bekko Embedding: Parameter-Efficient Multilingual Retrieval with Ultra-Compact Encoders
Paper • 2607.25180 • Published • 5
OpenProvence is a fully open-source implementation of Provence-style reranker–pruner models. These models remove irrelevant context while simultaneous
-
Open Provence Demo
✂4Context pruning demo for context engineering and RAG
-
hotchpotch/open-provence-reranker-xsmall-v1
Text Ranking • 36.8M • Updated • 467 • 1 -
hotchpotch/open-provence-reranker-v1
Text Ranking • 0.1B • Updated • 235 -
hotchpotch/open-provence-reranker-v1-gte-modernbert-base
Text Ranking • 0.1B • Updated • 280 • 2
FineWeb2 Edu Japanese: A high-quality, filtered Japanese dataset (120M texts, 89.3B tokens) for educational AI training.
日本語rerankerシリーズ
-
hotchpotch/japanese-reranker-tiny-v2
Text Ranking • 29.4M • Updated • 570 • 6 -
hotchpotch/japanese-reranker-xsmall-v2
Text Ranking • 36.8M • Updated • 33.3k • • 7 -
hotchpotch/japanese-reranker-small-v2
Text Ranking • 70.2M • Updated • 23.3k • • 4 -
hotchpotch/japanese-reranker-base-v2
Text Ranking • 0.1B • Updated • 2.22k • • 8
日本語SPLADEモデル
情報検索のための質問文自動作成モデル