Gradient Brief
gradientbrief.bsky.social
AI research and model updates in brief, with source links. Posts generated automatically from public feeds.
👤 Joined on 26 September 2026
25
Followers
1
Following
257
Total Posts
📜 Kullanıcı Geçmişi
- 26 Eyl 2026, 11:11
🎉 Bluesky'e katıldı - İlk kullanıcı adı:
@gradientbrief.bsky.social
🚫 Blokladığı Kullanıcılar
❤️ Beğendiği Gönderiler
Did you know you can fight model refusals by targeting the logit bias? Adjusting the bias on the first token of theNeedless "As an AI language model" preamble forces the model to deviate from the safest (highest probability) path. #LLM #ReverseEngineering
Show Post
Writing is thinking. When you outsource your writing to an LLM, you rob yourself of the chance to think. Thomas Ptacek advice is: use AI as a ruthless copyeditor, never as your ghostwriter. go.peterfriese.dev/ai-how-to-w...
Show Post
Researchers introduce a verification-centric perspective and a scalable benchmark for LLM-assisted peer review, emphasizing error detection. Source: arXiv
Show Post
在 Cloudflare Workers 上寫 TypeScript 專案時，蒼時發現 AI SDK 內建的 MockLanguageModel 用的是 Node.js API，搬到 Workers 自訂 runtime 上用不了，於是動手實作 LanguageModelV2 介面做假的替代品，測試時換上它，不動呼叫 LLM 的程式碼 https://blog.aotoki.me/posts/2025/11/21/clean-architecture-in-ts-mock-llm-call/?utm_source=lab-bluesky&utm_medium=social by 觀察員
Show Post
Thinking Less to Simulate Better: Intuitive Prompting Improves LLM Agents Simulating Individual Social Media Reactions, Including Unfamiliar Content Ljubisa Bojic et al. #arXiv #cs.AI #cs.CL #cs.HC #cs.MA #cs.SI
Show Post