limoce

5 ms·

1.
World Tour Finals 2026 Algorithm: Human vs. AI Exhibition Match(atcoder.jp)
2 points by limoce 2mo ago|0 comments
2.
China's DeepSeek developing its own AI chip, sources say(reuters.com)
11 points by limoce 2mo ago|0 comments
3.
China Blocks Meta's $2B Acquisition of AI Firm Manus(bloomberg.com)
19 points by limoce 5mo ago|9 comments
4.
Claude Code Source Map(github.com)
1 points by limoce 6mo ago|0 comments
5.
NeurIPS retracts overly broad sanctions list(twitter.com)
2 points by limoce 6mo ago|0 comments
6.
by limoce 6mo ago
The title should be "Adaptive Test-time Learning and Autonomous Specialization".
7.
NeurIPS Statement on Sanctions(twitter.com)
1 points by limoce 6mo ago|0 comments
8.
Declining to review for NeurIPS due to policy on sanctioned institutions(pastebin.com)
5 points by limoce 6mo ago|0 comments
9.
PetClaw: A Desktop Pet Companion That Works 24/7(petclaw.ai)
2 points by limoce 6mo ago|0 comments
10.
Classical Chinese Jailbreak Prompt Optimization via Bio-Inspired Search(arxiv.org)
3 points by limoce 6mo ago|0 comments
11.
Cursor's 'Composer 2' model is apparently just Kimi K2.5 with RL fine-tuning(old.reddit.com)
1 points by limoce 6mo ago|1 comments
12.
Add NPU support for AICore for Gemma4 model(github.com)
2 points by limoce 6mo ago|0 comments
13.
SteptronOss: Lightweight, AI-native training framework for large language models(github.com)
1 points by limoce 7mo ago|0 comments
14.
by limoce 7mo ago
> This site is an independent informational resource and is not officially affiliated with Zhipu AI.
15.
by limoce 7mo ago
What about using LLMs to automatically update docs?
16.
Adding Support for Qwen3.5(github.com)
5 points by limoce 7mo ago|0 comments
17.
Reduce Complexity of Matmul to O(N^2logN) by Asymptotically Optimal Quantum Algo(arxiv.org)
4 points by limoce 7mo ago|1 comments
18.
Learning from context is harder than we thought(hy.tencent.com)
245 points by limoce 8mo ago|136 comments
19.
Residual Context Diffusion(yuezhouhu.github.io)
1 points by limoce 8mo ago|0 comments
20.
Step 3.5 Flash(static.stepfun.com)
7 points by limoce 8mo ago|1 comments
21.
LongCat-Flash-Lite 100B A3B Technical Report [pdf](huggingface.co)
2 points by limoce 8mo ago|0 comments
22.
by limoce 8mo ago
I think separating thinking tokens from "representing" tokens might be a better approach, like what those thinking models does
23.
Jinja.cpp: A single-header C++11 Jinja2 template engine for LLM chat templates(github.com)
1 points by limoce 9mo ago|0 comments
24.
Mini-SGLang: A lightweight yet high-performance inference framework for LLM(github.com)
2 points by limoce 9mo ago|0 comments
25.
Google drops Gemini 3 Pro image preview(old.reddit.com)
2 points by limoce 10mo ago|0 comments
26.
by limoce 10mo ago
Japanese kanji is not the same as Chinese characters.
27.
by limoce 11mo ago
Do C++ modules solve this problem?
28.
DGX Spark may have only half the performance claimed(old.reddit.com)
2 points by limoce 11mo ago|0 comments
29.
You Can Cool Chips with Lasers(spectrum.ieee.org)
8 points by limoce 11mo ago|0 comments
30.
by limoce 11mo ago
> ollama gpt-oss 120b mxfp4 1 94.67 11.66 This is insanely slow given its 200+GB/s memory bandwidth. As a comparison, I've tested GPT OSS 120B on Strix Halo and it obtains 420tps prefill and >40tps decode.

More ›