Hacker News: at2005

Matrix Orthogonalization Improves Memory in Recurrent Models

at2005 — Wed, 01 Jul 2026 05:13:22 +0000

Article URL: https://ayushtambde.com/blog/matrix-orthogonalization-improves-memory-in-recurrent-models/

Comments URL: https://news.ycombinator.com/item?id=48742514

Points: 85

# Comments: 32

New comment by at2005 in "Tree Search Distillation for Language Models Using PPO"

at2005 — Mon, 16 Mar 2026 21:47:19 +0000

I didn't compare with the harness (focused on distillation) but the original ToT paper has a section on it: https://arxiv.org/abs/2305.10601

New comment by at2005 in "Tree Search Distillation for Language Models Using PPO"

at2005 — Sun, 15 Mar 2026 06:50:01 +0000

Ah, I meant that MCTS uses more inference-time compute (over GRPO) to produce a training sample

Tree Search Distillation for Language Models Using PPO

at2005 — Sun, 15 Mar 2026 00:51:55 +0000

Article URL: https://ayushtambde.com/blog/tree-search-distillation-for-language-models-using-ppo/

Comments URL: https://news.ycombinator.com/item?id=47383059

Points: 87

# Comments: 10

New comment by at2005 in "A HL Programming Language for Quantum Computers"

at2005 — Tue, 23 Feb 2021 19:34:17 +0000

Btw the whole motivation for this were algorithms like Grover's, which need "oracles" to be specified. You can only imagine trying to code adders and greater-than circuits with QASM...

A HL Programming Language for Quantum Computers

at2005 — Tue, 23 Feb 2021 19:29:15 +0000

Hi, I've designed and implemented a higher-level programming language for quantum computers. It essentially makes creating classical oracles for use in quantum circuits much easier, by defining C-style arithmetic/logical operations on qregs and having a compiler pass to estimate qubit counts prior to compilation, which, btw compiles to QASM for use on IBM systems. Does this seem interesting to people? Would anyone like to learn more about it?

Comments URL: https://news.ycombinator.com/item?id=26241699

Points: 1

# Comments: 1