Russian User Memory Benchmark: Evaluating long-context memory
ai-forever PRO
ai-forever
AI & ML interests
NLP, CV, multimodal, python3, pytorch, machine learning, deep learning
Recent Activity
published a dataset 1 day ago
ai-forever/RUMBA-leaderboard-data updated a Space 1 day ago
ai-forever/RUMBA-leaderboard published a dataset 1 day ago
ai-forever/RUMBAOrganizations
-
kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s-Diffusers
2B • Updated • 173 • 4 -
kandinskylab/Kandinsky-5.0-T2V-Lite-nocfg-5s-Diffusers
2B • Updated • 25 • 1 -
kandinskylab/Kandinsky-5.0-T2V-Lite-distilled16steps-5s-Diffusers
2B • Updated • 62 • 1 -
kandinskylab/Kandinsky-5.0-T2V-Lite-pretrain-5s-Diffusers
2B • Updated • 1 • 1
Collection of finetuned models for NABLA: Neighborhood Adaptive Block-Level Attention
https://arxiv.org/abs/2507.13546
-
ai-forever/Wan2.1-T2V-14B-NABLA-0.7
Text-to-Video • 14B • Updated • 299 • 6 -
ai-forever/Wan2.1-T2V-14B-NABLA-0.5-STA-11-5-5
Text-to-Video • 14B • Updated • 115 -
ai-forever/Wan2.1-T2V-14B-NABLA-0.6-STA-11-3-3
Text-to-Video • 14B • Updated • 84 • 1 -
nablaNABLA: Neighborhood Adaptive Block-Level Attention
Paper • 2507.13546 • Published • 126
FiMMIA: scaling semantic perturbation-based membership inference across modalities.
LLM-as-a-judge
-
ai-forever/pollux-judge-7b
Text Generation • 8B • Updated • 459 • • 2 -
ai-forever/pollux-judge-7b-r
Text Generation • 8B • Updated • 10 • 4 -
ai-forever/pollux-judge-32b
Text Generation • 33B • Updated • 65 • • 6 -
ai-forever/pollux-judge-32b-r
Text Generation • 33B • Updated • 17 • 1
-
ai-forever/georeview-classification
Updated • 66 -
ai-forever/headline-classification
Viewer • Updated • 60k • 364 • 1 -
ai-forever/inappropriateness-classification
Viewer • Updated • 54k • 159 • 1 -
ai-forever/kinopoisk-sentiment-classification
Viewer • Updated • 13.5k • 355 • 6
Collection with our sentence embedding models, like sbert, e5 etc.
-
ai-forever/sbert_large_mt_nlu_ru
Feature Extraction • 0.4B • Updated • 1.91k • • 26 -
ai-forever/sbert_large_nlu_ru
Feature Extraction • 0.4B • Updated • 25.3k • • 104 -
ai-forever/ru-en-RoSBERTa
Feature Extraction • 0.4B • Updated • 41.9k • • 85 -
ai-forever/FRIDA
Feature Extraction • 0.8B • Updated • 178k • • 146
Russian User Memory Benchmark: Evaluating long-context memory
FiMMIA: scaling semantic perturbation-based membership inference across modalities.
-
kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s-Diffusers
2B • Updated • 173 • 4 -
kandinskylab/Kandinsky-5.0-T2V-Lite-nocfg-5s-Diffusers
2B • Updated • 25 • 1 -
kandinskylab/Kandinsky-5.0-T2V-Lite-distilled16steps-5s-Diffusers
2B • Updated • 62 • 1 -
kandinskylab/Kandinsky-5.0-T2V-Lite-pretrain-5s-Diffusers
2B • Updated • 1 • 1
Collection of finetuned models for NABLA: Neighborhood Adaptive Block-Level Attention
https://arxiv.org/abs/2507.13546
-
ai-forever/Wan2.1-T2V-14B-NABLA-0.7
Text-to-Video • 14B • Updated • 299 • 6 -
ai-forever/Wan2.1-T2V-14B-NABLA-0.5-STA-11-5-5
Text-to-Video • 14B • Updated • 115 -
ai-forever/Wan2.1-T2V-14B-NABLA-0.6-STA-11-3-3
Text-to-Video • 14B • Updated • 84 • 1 -
nablaNABLA: Neighborhood Adaptive Block-Level Attention
Paper • 2507.13546 • Published • 126
LLM-as-a-judge
-
ai-forever/pollux-judge-7b
Text Generation • 8B • Updated • 459 • • 2 -
ai-forever/pollux-judge-7b-r
Text Generation • 8B • Updated • 10 • 4 -
ai-forever/pollux-judge-32b
Text Generation • 33B • Updated • 65 • • 6 -
ai-forever/pollux-judge-32b-r
Text Generation • 33B • Updated • 17 • 1
-
ai-forever/georeview-classification
Updated • 66 -
ai-forever/headline-classification
Viewer • Updated • 60k • 364 • 1 -
ai-forever/inappropriateness-classification
Viewer • Updated • 54k • 159 • 1 -
ai-forever/kinopoisk-sentiment-classification
Viewer • Updated • 13.5k • 355 • 6
Collection with our sentence embedding models, like sbert, e5 etc.
-
ai-forever/sbert_large_mt_nlu_ru
Feature Extraction • 0.4B • Updated • 1.91k • • 26 -
ai-forever/sbert_large_nlu_ru
Feature Extraction • 0.4B • Updated • 25.3k • • 104 -
ai-forever/ru-en-RoSBERTa
Feature Extraction • 0.4B • Updated • 41.9k • • 85 -
ai-forever/FRIDA
Feature Extraction • 0.8B • Updated • 178k • • 146