The models trained under our Karasu and Qarasu project
Lightblue KK.
company
AI & ML interests
None defined yet.
Recent Activity
View all activity
Organization Card
We are a Tokyo-based firm striving to make the most useful, trustworthy, and reliable AI for our customers and the AI community.
Find out more about us on our website.
Collections
6
models
55
lightblue/Karasu-DPO-7B
Text Generation
•
Updated
lightblue/lb-reranker-0.5B-v1.0
Text Generation
•
Updated
•
2.01k
•
60
lightblue/reranker_0.5_cont_filt_7max
Text Generation
•
Updated
•
137
lightblue/reranker_0.5_bincont_filt
Text Generation
•
Updated
•
122
lightblue/reranker_0.5_cont
Text Generation
•
Updated
•
121
lightblue/reranker_0.5_bin_filt
Text Generation
•
Updated
•
129
•
1
lightblue/reranker_0.5_cont_filt
Text Generation
•
Updated
•
119
lightblue/qwen2.5-7B-instruct-simpo
Text Generation
•
Updated
•
12
lightblue/qwen2.5-7B-instruct-orpo2
Text Generation
•
Updated
•
7
lightblue/qwen2.5-7B-instruct-orpo
Text Generation
•
Updated
•
13
•
1
datasets
28
lightblue/distilabel-reasoning-R1-Llama-70B
Viewer
•
Updated
•
2.43k
•
21
lightblue/rag_multilingual_training_negatives
Viewer
•
Updated
•
220k
•
380
•
2
lightblue/Sky-T1_data_17k_ja
Viewer
•
Updated
•
16.4k
•
12
lightblue/reranker_continuous_filt_max7_train
Viewer
•
Updated
•
2.28M
•
114
•
3
lightblue/rag_datasets_selected_32B4scored_probs
Viewer
•
Updated
•
2.55M
•
55
lightblue/rag_datasets_selected_32B4scored
Viewer
•
Updated
•
2.55M
•
44
lightblue/rag_datasets_selected_14Bscored
Viewer
•
Updated
•
2.66M
•
44
lightblue/response-dataset-plus-qwen-judged
Viewer
•
Updated
•
50.8k
•
49
•
1
lightblue/response-dataset-plus-qwen
Viewer
•
Updated
•
50.8k
•
45
lightblue/reranker_fulltrain_scored_filtered
Viewer
•
Updated
•
2.58M
•
42