Skip to main content
SandRise logo SandRise
Exploring Next / By company / Deepseek V4

Topic

Deepseek V4

3 episodes

  1. Ep 729 Jul 21, 2026

    Overview: Active vs Total Parameters

    We finally slow down on active vs total parameters, because we keep throwing the phrase around like it explains itself. This is us making the difference click: what a model stores versus what it actually uses when it answers.

    New ModelsInferenceTencent Hy3Deepseek V4
  2. Ep 470 Jun 9, 2026

    FlashMemory DeepSeek V4: Lightning Index Ultra Long Context via Lookahead Sparse Attention

    Researchers propose Lookahead Sparse Attention (LSA) with a Neural Memory Indexer to slash GPU memory usage for ultra-long LLM context by pre-predicting which KV cache chunks matter, trained independently without the full backbone. FlashMemory-DeepSeek-V4 cuts physical KV cache to 13.5% of baseline on average while maintaining or improving accuracy (+0.6% abs) across LongBench-v2, LongMemEval, RULER—at 500K tokens, it suppresses KV overhead by over 90%. Project paused due to org changes; code not yet public.

    InferenceNew ModelsDeepseek V4Flashmemory
  3. Ep 466 Jun 5, 2026

    Rethinking Continual Experience Internalization for Self Evolving LLM Agents

    Jingwen Chen et al. diagnose why iterative experience internalization fails in LLMs and prescribe a three-part fix—principle-level granularity, step-wise injection, off-policy context-distillation—that turns capability collapse into compounding improvement.

    AgentsTrainingQwen3Deepseek V4
SandRise logo SandRise Product Studio
Resume LinkedIn GitHub Email

© 2026 SandRise · Built by Nick Sanders

🧠 PM Perspective

Crafting your PM challenge
Analyzing context and generating a thoughtful question...
Your Challenge
0 / 2000
✨

Feedback on Your Answer

⚠️

Say Hi

Feedback, ideas, interesting finds — anything goes.

What's this about?
0 / 2,000

Note received!

Thanks for reaching out. I'll take a look soon.

⚠️

Something went wrong. Please try again.