Видео ютуба по тегу Token Compression

Data-Centric LLM Token Compression
Data-Centric LLM Token Compression
UniPruneBench: Visual Token Compression Benchmark
UniPruneBench: Visual Token Compression Benchmark
LLaVA-Scissor: Semantic Video Compression
LLaVA-Scissor: Semantic Video Compression
CompressX - Solana Token Compression Service (Demo)
CompressX - Solana Token Compression Service (Demo)
LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs
LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs
Token-Efficient Long Video Understanding for Multimodal LLMs | Paper explained
Token-Efficient Long Video Understanding for Multimodal LLMs | Paper explained
[QA] LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs
[QA] LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs
4.3.2. Prompt Compression Explained
4.3.2. Prompt Compression Explained
Token Cost Reduction through LLMLingua's Prompt Compression
Token Cost Reduction through LLMLingua's Prompt Compression
Sequential Text Compression Strategy for LLMs | 65% Token Reduction with NLP + Data Cleaning
Sequential Text Compression Strategy for LLMs | 65% Token Reduction with NLP + Data Cleaning
MaGMML3 - Interspatial Compression (Token Taker) - 1 Token in 1:36.966 (WR)
MaGMML3 - Interspatial Compression (Token Taker) - 1 Token in 1:36.966 (WR)
4.3.2. Prompt Compression Explained
4.3.2. Prompt Compression Explained
2510.18234 - DeepSeek OCR: Contexts Optical Compression
2510.18234 - DeepSeek OCR: Contexts Optical Compression
Bridging the Gap: Encoding Long Text with Highly Compressed Yet Learnable Tokens
Bridging the Gap: Encoding Long Text with Highly Compressed Yet Learnable Tokens
Style compression: Make LLMs sound less generic (with fewer tokens)
Style compression: Make LLMs sound less generic (with fewer tokens)
[2025 Team 3] Hierarchical Token Compression for Video Summarization with MLLMs
[2025 Team 3] Hierarchical Token Compression for Video Summarization with MLLMs
Объяснение DeepSeek-OCR
Объяснение DeepSeek-OCR
Let's build the GPT Tokenizer
Let's build the GPT Tokenizer
DeepSeek-OCR Explained: How Visual Tokens Compress Text for Efficient LLMs
DeepSeek-OCR Explained: How Visual Tokens Compress Text for Efficient LLMs
[QA] TokenSkip: Controllable Chain-of-Thought Compression in LLMs
[QA] TokenSkip: Controllable Chain-of-Thought Compression in LLMs