INDEX

Explanations

words related to obtaining something, especially through legal or official means

oai_token-act-pair · gpt-3.5-turbo

instances of the word "obtain" indicating processes of acquisition or permission

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

psc

-0.72

hov

-0.71

timer

-0.66

heartedly

-0.66

dr

-0.66

 Wells

-0.63

cas

-0.63

amd

-0.61

gaard

-0.61

pl

-0.60

POSITIVE LOGITS

 access

0.87

 permission

0.84

å¥³

0.82

 signatures

0.81

itures

0.79

 authorization

0.79

andom

0.77

 permits

0.75

 licenses

0.74

 warrants

0.73

Activations Density 0.042%

words related to obtaining something, especially through legal or official means

instances of the word "obtain" indicating processes of acquisition or permission

No Comments

No Known Activations