INDEX

Explanations

concepts related to comparison and order among items or entities

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_10.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.10.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Negative Logits

itans

-0.56

ortium

-0.54

apo

-0.54

(>

-0.54

enegger

-0.54

poon

-0.53

umar

-0.53

ullivan

-0.53

yout

-0.52

pected

-0.51

POSITIVE LOGITS

 differently

1.83

 different

1.81

different

1.69

 similar

1.42

 opposite

1.41

 worse

1.35

 Different

1.35

 identical

1.32

 similarly

1.26

 differing

1.23

Activations Density 5.132%

concepts related to comparison and order among items or entities

No Known Activations