AI for Scientific Search(arxiv.org)
108 points by omarsar | 26 comments
First screen, captured .
Looks similar
Efficient Reasoning with Hidden Thinking(arxiv.org)
152 points by fofoz | 37 comments
DeepSeek-V3 Technical Report(arxiv.org)
107 points by signa11 | 27 comments
Related stories
Ladder: Self-improving LLMs through recursive problem decomposition(arxiv.org)
339 points by fofoz | 103 comments
arXiv paper introducing LADDER, a framework where LLMs self-improve by recursively generating simpler problem variants, plus TTRL test-time reinforcement learning.
DeepSeek reasonix, DeepSeek native coding agent with high caching and low cost(esengine.github.io)
706 points by Alifatisk | 273 comments
Project homepage for Reasonix, an open-source DeepSeek-native terminal coding agent built around DeepSeek's prefix cache for low-cost long sessions.
LIMO: Less Is More for Reasoning(arxiv.org)
376 points by trott | 126 comments
arXiv paper presenting LIMO, a model showing complex math reasoning in LLMs can be elicited with only 817 curated training samples.
Understanding Reasoning LLMs(magazine.sebastianraschka.com)
442 points by sebg | 163 comments
Sebastian Raschka's Ahead of AI newsletter article explaining the four main approaches to building and refining reasoning LLMs.
There Will Be a Scientific Theory of Deep Learning(arxiv.org)
320 points by jamie-simon | 142 comments
An arXiv position paper arguing that a scientific theory of deep learning is emerging, surveying five strands of research such as solvable models, scaling limits, and universal behaviors.
Stealing Reasoning Traces from Proprietary LLM APIs(stolen-thoughts.com)
665 points by quantumgarbage | 295 comments
Project page for a research paper on extracting hidden reasoning traces from proprietary LLM APIs, by authors from MATS, ELLIS, Max Planck and Tübingen.