INDEX

Explanations

phrases related to news or headlines

oai_token-act-pair · gpt-3.5-turbo

occurrences of the abbreviation "NE."

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

gerald

-0.65

stroke

-0.64

 pains

-0.64

kaya

-0.63

trap

-0.61

 Hussein

-0.60

gradient

-0.60

 totality

-0.60

 declass

-0.59

acies

-0.59

POSITIVE LOGITS

ITH

1.00

braska

0.95

VE

0.92

NE

0.91

IGH

0.84

erd

0.82

VER

0.81

JM

0.81

LL

0.80

ISS

0.80

Activations Density 0.006%

phrases related to news or headlines

occurrences of the abbreviation "NE."

No Comments

No Known Activations