INDEX

Explanations

phrases or words related to different ways of doing something

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate various methods or approaches

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

uster

-0.72

asts

-0.71

isher

-0.61

icio

-0.60

ãĥ¡

-0.60

oak

-0.60

akov

-0.58

Ä±

-0.58

usters

-0.58

arthed

-0.58

POSITIVE LOGITS

 ways

1.10

finding

0.99

 Ways

0.87

isms

0.82

point

0.80

terday

0.77

styles

0.77

pointers

0.74

forward

0.73

steps

0.71

Activations Density 0.015%

phrases or words related to different ways of doing something

phrases that indicate various methods or approaches

No Comments

No Known Activations