xianshou

2,816 karmajoined May 1, 2011On HN

I live in New York, work in finance, drink excessive amounts of coffee, and play chess.

Recent submissions

Merge and Conquer: Evolutionarily Optimizing AI for 2048 arxiv.org
1xianshou10modiscuss
Stuck in the Matrix: Probing Spatial Reasoning in Large Language Models arxiv.org
1xianshou10modiscuss
Reflection AI Raises $2B to Build "American DeepSeek"nytimes.com
9xianshou11mo2 comments
Nvidia-backed Reflection AI raising at $5.5B valuation reuters.com
2xianshou11mo1 comment
Unsupervised Elicitation of Language Models arxiv.org
7xianshou1ydiscuss
DeepSeek V3 0324 is now the best nonthinking model (Reddit)old.reddit.com
1xianshou1ydiscuss
DeepSeek V3 0324 outpaces GPT 4.5 and Claude 3.7 in coding, other benchmarks huggingface.co
7xianshou1ydiscuss
Practical RL (Yandex Data School)github.com/yandexdataschool
1xianshou1ydiscuss
InvestorBench: A Benchmark for Financial Decision-Making Tasks with Agents arxiv.org
1xianshou1ydiscuss
An Evolved Universal Transformer Memory (Sakana.ai)sakana.ai
1xianshou1ydiscuss

All submissions and comments on Hacker News.