INDEX

Explanations

specific instances of TA followed by a number

oai_token-act-pair · gpt-3.5-turbo

references to "TA" in various contexts, likely indicating a specific term or acronym

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

space

-0.75

meyer

-0.71

ician

-0.71

nen

-0.70

builders

-0.69

comes

-0.69

papers

-0.68

sburg

-0.68

bah

-0.68

nan

-0.66

POSITIVE LOGITS

KE

1.05

VE

0.96

WA

0.94

WN

0.91

BILITY

0.91

FU

0.89

KING

0.89

ZE

0.87

BIL

0.87

KER

0.87

Activations Density 0.008%

specific instances of TA followed by a number

references to "TA" in various contexts, likely indicating a specific term or acronym

No Comments

No Known Activations