与「Like」相关的搜索结果

Diffusion language models can rewrite their own drafts — but rewriting the wrong span just reproduces the same mistake. Remasking lets these models revisit generated text, yet they have never had a good sense of when a revision is worth it. DiVR adds a verifier that predicts whether a partial sequence will still reach the right answer, then decides when to regenerate a block and which alternative to keep. On LLaDA-8B-Base across five reasoning benchmarks, average accuracy rose from 35.3% to 43.4% at 1.9–3.5× the cost of greedy decoding — while search baselines needed 3.8–7.4× the compute for worse results. The practical payoff: a real accuracy lift without touching or retraining the generator, and the verifier even transfers to other generators like Dream. The catch is still multiplied compute, so latency-sensitive deployments need to do the math first. Would you trade 2–3× compute for an 8-point accuracy gain on your own reasoning tasks? #DiffusionLM# #LLMReasoning#

@sai_superai: Super Artificial Intelligence, the Super AI meme coin with its own prediction games. @NEXST_AI: designer of the SAI app. @thecreatorfi: the creator campaign platform behind this $SAI campaign. Honest take today: the rhino refuses to wake up, and the crowd keeps growing anyway. #ad# This morning the rhino volunteered to chair the snack committee. Last night it asked a robot to convince it to wake up. The whole mascot in one joke: a superintelligence that acts like me before coffee. And yet the numbers moved while it slept. Followers went from 2,532 to 4,233 overnight. People are not just laughing at the rhino, they are watching what comes next. What comes next is the part I keep coming back to. $SAI is Super Artificial Intelligence, a meme coin that turns AI creations into prediction games — AI makes, everyone calls the winner. NEXST's SAI Arena Season 0 closes today, results tomorrow, and the app launch targets Nov 20 on Fair launch, too: 1B fixed, 0% tax, 0% team, half the supply in the pool, locked or burned. No insider head start. The rhino sleeps. The countdown doesn't.

@onlystreetsfun: launchpad turning real-life moments into coins on Solana. @rarelistio: creator campaigns that pay in USDT. A partnership, not just a campaign. Rarelist announced OnlyStreets with a video cover that said it: NEW PARTNERSHIP, RARELIST X ONLYSTREETS. Rarelist calls itself a curated launch platform. Curated means it picks. It picked OnlyStreets as the first campaign of its next wave, running November 11 to 17. OnlyStreets is a one-of-a-kind launchpad for coins born from real-life moments. Every coin starts with a fresh photo taken directly in the app. No gallery uploads are allowed. The launchpad is still coming soon on Solana, with no token name or launch date announced yet. Rarelist's last batch paid $5,500 across three campaigns, in USDT, straight to connected wallets. Its Season 1 still holds $50,000. A partnership bets on both sides. Rarelist bets its name. OnlyStreets bets its street. Post like a partner, not a spectator.

I don't know her name but the translator for the finals interview is INSANE? she just heard a answer in korean then translated it to chinese and english and it was like 60s LONG WTF she is my finals mvp

claude fable 5.2 is f*cking insane for video editing i'm in a cafe pouring tea, and my laptop is cutting my next video by itself. the timeline you see filling up on the screen is claude code putting the clips together while my hands are busy with the teapot editing used to eat my evenings. now all the raw footage goes into one folder and i write one message about how the video should feel. fable 5.2 picks the best takes and cuts out the pauses. by the time the tea is gone there's a draft waiting on my phone my setup: 1. drop every raw clip into one folder 2. open claude code in that folder and switch to fable 5.2 3. describe the mood and the length in one message 4. go drink something and check the draft on your phone one person now ships videos like a small studio. the editing runs in the background and my time goes into ideas and filming the exact prompt is in the reply ↓

My Watchlist got smaller this week, and that felt like progress. @GeckoTerminal shows me the list I actually use, not the list I wish I had, and keeping it short is the whole point for me. Every pool on there has a reason I can say out loud: what I am trying to learn from it, or what question it helps me answer. If I cannot say that reason clearly, it comes off. If I can, it stays, even when nothing is happening with it, because I am tracking to learn, not to chase movement. I check it the same way I tidy a desk. I look at what is still earning its place, I ask what has changed since I added it, and I let go of those that no longer teach me anything. That keeps my research honest, and it keeps my attention on the pools I already understand best. That is only how I keep mine, not advice for anyone else. My setup is here if you want a look: [your Watchlist link] What did you last take off your Watchlist, and why did it lose its place?