MIT and MIT-IBM researchers' 'Swarm Telepathy (SwaT)' lets models from different companies share KV caches directly instead of text, claiming 40% higher accuracy and 4x speed (report)

Verified (fact)

Researchers from MIT CSAIL and the MIT-IBM lab posted a paper titled 'Swarm Telepathy: Prefill-Free KV Cache Transfer Across Model Families' to an online archive, AI Times reported. Instead of exchanging text, one model passes its KV cache, the result of its internal computation, directly to another model. The method reportedly works even between models made by different companies.

Needs fact-check (claim)

The researchers claim average accuracy is 40% higher and processing is 4x faster than text-based communication. The report does not say which tasks or model combinations these figures were measured on.

Why it matters (interpretation)

In multi-agent systems, prefill costs pile up every time models exchange text. If this approach works across different model families, it could significantly cut the cost and latency of agent collaboration.

Sources

• Press MIT, AI 모델끼리 글 없이 소통하는 ‘스웜 텔레파시’ 개발 AI Times (Korea)

More about MIT

• 2026-11-03 P2 Research Paper Language modelsSafety & alignment MITOpenAIAlibaba
  Weight-Space Programming follow-up: critics say the Qwen improvement relied on Python execution; partly reproduced with Qwen 3.8 27B on a laptop (community reports)

  A rebuttal has emerged to the paper "Sparks of Weight-Space Programming" (authors: Rikiya Takehi, Jayden Teoh, Alex Zhang, Omar...

• 2026-11-03 P1 Research Paper Language modelsSafety & alignmentEfficiency & infra MITOpenAIAlibaba
  GPT-6 Astra reportedly edits just 0.016% of Qwen weights without training, lifting MATH-500 from 7.9% to 48.1% (community report)

  A research blog post, 'Sparks of Weight-Space Programming,' published early experimental results with GPT-6 Astra, saying it can...

• 2026-10-30 P2 Open source Open weights Generative mediaComputer vision NVIDIAMiniMaxMIT
  NVIDIA and MIT release weights for 'Open-H3-Regenerate-2K', a video enhancement model built on MiniMax H3 that regenerates up to 4K (report)

  According to QbitAI (量子位), NVIDIA and MIT researchers uploaded the model weights for the video enhancement project...

• 2026-10-26 P3 Research Paper AgentsSafety & alignment MITAnthropic
  Follow-up to virtual-island Claude agent experiment: MIT paper 'Is this machine playing?' finds a throwing technique unknown to the developers (reported)

  MIT researchers posted a paper on arXiv, 'Is this machine playing?', analyzing how AI agents released on a virtual island without any...

• 2026-10-18 P2 Research Paper Science & bioMultimodal MIT
  MIT researchers develop a multimodal model that understands 3D atomic arrangements together with natural language, applied to solid-state battery electrolyte discovery (report)

  According to AI Shinmun, MIT researchers have built a multimodal model that learns by linking crystalline materials' atomic arrangements...