@vosje62@mastodon.nl2026-01-05 20:39:34
Ofcom makes 'urgent contact' with X over concerns Grok AI can generate 'sexualised images of children' | Science, Climate & Tech News | Sky News
#Twitter
@cosmos4u@scicomm.xyz2026-02-03 17:52:08
In 10 minutes the #ArtemisII Fueling Test News Conference will stream at https://www.youtube.com/live/ycqk3uN_N6g - as announced in https://www.nasa.gov/blogs/missions/2026/02/03/nasa-conducts-artemis-ii-fuel-test-eyes-march-for-launch-opportunity/ and https://x.com/NASAAdmin/status/2018578937115271660 the first launch attempt has already been moved to March.
@sncf_ligne_r@lepoulsdumonde.com2026-02-04 17:26:49
Prévoir un allongement de temps de trajet de 15 minutes environ pour le train suivant : - KUMO, départ Paris Gare de Lyon 17h44 - arrivée Montereau 18h44
Pour plus d'informations sur cette perturbation, consultez le fil X de la ligne.
Motif : alerte de sécurité émise par le conducteur.
🤖 04/02 18:26
@raiders@darktundra.xyz2026-02-05 00:57:33
Former Raiders HC Jon Gruden Issued Compelling Jets Prediction https://heavy.com/sports/nfl/las-vegas-raiders/jon-gruden-compelling-jets-prediction/
@Techmeme@techhub.social2026-02-24 15:30:54
Kraken launches what it claims are the first regulated perpetual futures contracts based on tokenized stocks, available to non-US users and trading 24/7 (Krisztian Sandor/CoinDesk)
https://www.coindesk.com/business/2026/02/
@rer_c@lepoulsdumonde.com2026-02-04 11:58:51
Le trafic est perturbé entre Versailles Chantiers et Massy-Palaiseau.
Des suppressions, ralentissements ou modifications de dessertes sont Š prévoir.
Pour plus d'informations sur cette perturbation, consultez le fil X du RER C.
Motif : difficulté d'acheminement du conducteur (quelques trains supprimés).
🤖 04/02 12:58
@arXiv_csOS_bot@mastoxiv.page2026-02-04 07:41:57
ProphetKV: User-Query-Driven Selective Recomputation for Efficient KV Cache Reuse in Retrieval-Augmented Generation
Shihao Wang, Jiahao Chen, Yanqi Pan, Hao Huang, Yichen Hao, Xiangyu Zou, Wen Xia, Wentao Zhang, Haitao Wang, Junhong Li, Chongyang Qiu, Pengfei Wang
https://arxiv.org/abs/2602.02579 https://arxiv.org/pdf/2602.02579 https://arxiv.org/html/2602.02579
arXiv:2602.02579v1 Announce Type: new
Abstract: The prefill stage of long-context Retrieval-Augmented Generation (RAG) is severely bottlenecked by computational overhead. To mitigate this, recent methods assemble pre-calculated KV caches of retrieved RAG documents (by a user query) and reprocess selected tokens to recover cross-attention between these pre-calculated KV caches. However, we identify a fundamental "crowding-out effect" in current token selection criteria: globally salient but user-query-irrelevant tokens saturate the limited recomputation budget, displacing the tokens truly essential for answering the user query and degrading inference accuracy.
We propose ProphetKV, a user-query-driven KV Cache reuse method for RAG scenarios. ProphetKV dynamically prioritizes tokens based on their semantic relevance to the user query and employs a dual-stage recomputation pipeline to fuse layer-wise attention metrics into a high-utility set. By ensuring the recomputation budget is dedicated to bridging the informational gap between retrieved context and the user query, ProphetKV achieves high-fidelity attention recovery with minimal overhead. Our extensive evaluation results show that ProphetKV retains 96%-101% of full-prefill accuracy with only a 20% recomputation ratio, while achieving accuracy improvements of 8.8%-24.9% on RULER and 18.6%-50.9% on LongBench over the state-of-the-art approaches (e.g., CacheBlend, EPIC, and KVShare).
toXiv_bot_toot
@leftsidestory@mstdn.social2026-01-29 02:47:52
Some City Some Nature VI 🏙️
一些城一些自然 VI 🏙️
📷 Zeiss IKON Super Ikonta 533/16
🎞️ Lucky SHD 400
#filmphotography #Photography #blackandwhite
@mcdanlj@social.makerforums.info2026-01-25 18:34:43
Working WFD CW today. Didn't yesterday due to doing storm preparations. Not a serious contester here, just pulled up not1mm, and am using it to send code so that I don't slow down more serious folks with my bad keying, just using the key when I need to send something that I didn't have a macro for, or to send slowly. Just now, I was delighted that I copied a 2x3 call on first hearing at 35WPM and then was deflated to hear "TEST" afterwards. I didn't know what contests are currently active, b…
@Techmeme@techhub.social2026-01-15 00:51:05
Thinking Machines parts ways with CTO Barret Zoph, with Soumith Chintala taking over the role; sources say the termination is due to "unethical conduct" (Kylie Robison/@kyliebytes)
https://x.com/kyliebytes/status/2011572331798548899
