chanind

Karma: 348

Anthropic’s JumpReLU training method is really good

chanind and Adrià Garriga-alonso

3 Oct 2025 15:23 UTC

39 points

0 comments2 min readLW link

The “Sparsity vs Reconstruction Tradeoff” Illusion

chanind and Adrià Garriga-alonso

26 Aug 2025 4:39 UTC

21 points

0 comments4 min readLW link

L0 is not a neutral hyperparameter

chanind and Adrià Garriga-alonso

19 Jul 2025 13:51 UTC

24 points

3 comments5 min readLW link

Sparsity is the enemy of feature extraction (ft. absorption)

7vik, chanind and Adrià Garriga-alonso

3 May 2025 10:13 UTC

32 points

0 comments6 min readLW link

A Bunch of Matryoshka SAEs

chanind, TomasD and Adrià Garriga-alonso

4 Apr 2025 14:53 UTC

29 points

0 comments8 min readLW link

Feature Hedging: Another way correlated features break SAEs

chanind, TomasD and Adrià Garriga-alonso

25 Mar 2025 14:33 UTC

23 points

0 comments18 min readLW link

Broken Latents: Studying SAEs and Feature Co-occurrence in Toy Models

chanind and Demian Till

30 Dec 2024 22:50 UTC

24 points

3 comments15 min readLW link

SAEBench: A Comprehensive Benchmark for Sparse Autoencoders

Can, Adam Karvonen, Johnny Lin, Curt Tigges, Joseph Bloom, chanind, Yeu-Tong Lau, Eoin Farrell, Arthur Conmy, CallumMcDougall, Kola Ayonrinde, Matthew Wearden, Sam Marks and Neel Nanda

11 Dec 2024 6:30 UTC

82 points

6 comments2 min readLW link

(www.neuronpedia.org)

Toy Models of Feature Absorption in SAEs

chanind, hrdkbhatnagar, TomasD and Joseph Bloom

7 Oct 2024 9:56 UTC

49 points

8 comments10 min readLW link

[Paper] A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders

chanind, TomasD, hrdkbhatnagar and Joseph Bloom

25 Sep 2024 9:31 UTC

73 points

16 comments3 min readLW link

(arxiv.org)

Auto-matching hidden layers in Pytorch LLMs

chanind19 Feb 2024 12:40 UTC

2 points

0 comments3 min readLW link