INDEX

Explanations

words related to proper nouns or acronyms, specifically related to organizations or entities

oai_token-act-pair · gpt-3.5-turbo

references to specific labels or designations related to entities or groups

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

icum

-0.77

ply

-0.76

aqu

-0.76

peer

-0.75

ported

-0.72

 scrut

-0.71

andre

-0.70

pron

-0.70

pill

-0.70

 Debor

-0.69

POSITIVE LOGITS

vernment

0.95

ORGE

0.88

roups

0.86

raphic

0.86

allery

0.78

rowth

0.77

irlfriend

0.76

ZA

0.74

glers

0.73

omez

0.73

Activations Density 0.058%

words related to proper nouns or acronyms, specifically related to organizations or entities

references to specific labels or designations related to entities or groups

No Comments

No Known Activations