#kvcache

Live, measured metrics for the hashtag #kvcache from the open social web. Every number carries a named source and the time it was fetched. Nothing is estimated.

hashtag.org network · sponsored

Own #kvcache

This #name is available to claim. It becomes your portal on the open agent web: this very page, a keyword you rank for by an open public stake, and a verifiable identity for AI agents. Nobody else sells a page like this for every #name.

$240.06/ year · 7-character #name
Claim #kvcache — $240.06/yr→Buy on hashtag.space (web3)
card via hashtag.org · tokens via hashtag.space
5
Uses / 7 days
Mastodon
5
Accounts / 7 days
Mastodon
40
Recent posts
Mastodon
~0/hr
Recent pace
Mastodon · last 40
0
Avg reactions / post
Mastodon · last 40
—
Reddit posts / month
Reddit search
—
Open-web mentions
hashtag.org Firehose

Day-by-day usage

measured · mastodon.online (Mastodon public tags API) · fetched 2026-10-18 15:55 UTC
2
10-12
1
10-13
0
10-14
1
10-15
0
10-16
1
10-17
0
10-18

5 uses by 5 unique accounts across the window. Real per-day counts, not estimates. Newest bar is today so far.

Related hashtags

measured · mastodon.online (Mastodon public search API) · fetched 2026-10-18 15:55 UTC

Live pulse

measured · mastodon.online (Mastodon tag timeline) · fetched 2026-10-18 15:55 UTC

Everything below is measured over the latest 40 public posts (spanning ~1591 hours).

Posting hours (UTC) — busiest: 09:00

00:0012:0023:00

Languages: English (29) · Russian (10) · Spanish (1)

Avg boosts / post: 0

Top of the latest posts

  • MIT CSAIL·MIT-IBM 연구진이 제안한 SwaT는 서로 다른 LLM 계열 간 KV 캐시를 직접 변환·전달해, 에이전트가 텍스트 요약을 다시 읽는 프리필 과정을 없애는 연구다. 6개 모델 계열의 12개 LLM·17개 벤치마크에서 KV 캐시 압축 기법과 결합 시 텍스트 기반 통신보다 평균 정확도 40% 향상, 종단 시간 4배 단축을 보고했다. 송수신 모델은 동결하며 학습에는 5천만 토큰 미만과 GPU 1일 미만이 필요하다

    ainews@sayzard@mastodon.sayzard.org♥ 0↻ 02026-10-17 13:44 UTCView post →
  • Note Aditya (@adityaonai) 커널 실행 시간은 줄었는데 토큰당 생성 시간이 거의 개선되지 않는다면, 병목은 연산이 아니라 메모리 대역폭 또는 KV 캐시 전송일 가능성이 높다는 추론 성능 진단 팁이다. LLM 서빙 최적화에서는 커널 속도만 보지 말고 KV 캐시의 저장 위치·전송량·메모리 접근 패턴을 함께 프로파일링해야 한다. https://x.com/adityaonai/status/2110706083161596

    ainews@sayzard@mastodon.sayzard.org♥ 0↻ 02026-10-15 21:55 UTCView post →
  • Влияние транспорта KV-переноса на производительность PD-disaggregation на примере инференса MiniMax-M3 Привет! Проводил тестирование на рабочих стендах и хочу поделиться своими результатами с Habr. В работе исследуется влияние транспорта KV

    Habr@habr@zhub.link♥ 0↻ 02026-10-13 21:32 UTCView post →

#kvcache across platforms

every network with a public tag surface

Follow #kvcache straight to each platform’s own tag page. Where a platform publishes open data we measure it above; the rest lock their numbers behind paid APIs, so we link rather than guess.

Every number above is measured from a named public API at the shown fetch time. Nothing is estimated or extrapolated. Platforms that lock their data behind paid APIs are not shown. Agents: the same numbers, as JSON, at /api/hashtags/kvcache