INDEX

Explanations

document in the digital world

oai_token-act-pair · gpt-3.5-turbo

references to "DL" and related identifiers or terms

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

borough

-0.89

tes

-0.85

ulas

-0.82

llor

-0.82

lio

-0.81

rament

-0.81

otle

-0.79

selage

-0.79

furt

-0.79

hell

-0.77

POSITIVE LOGITS

DL

0.86

anguage

0.77

DER

0.77

 tiss

0.73

ensor

0.71

DEN

0.70

 sling

0.70

ICT

0.70

DoS

0.70

DL

0.69

Activations Density 0.026%

document in the digital world

references to "DL" and related identifiers or terms

No Comments

No Known Activations