INDEX

Explanations

the acronym "NP" followed by a single-digit number

oai_token-act-pair · gpt-3.5-turbo

references to named entities or categories in a document

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hips

-0.82

ingham

-0.73

zona

-0.73

borgh

-0.71

iances

-0.69

mir

-0.69

âĸ¬

-0.68

bell

-0.68

hower

-0.67

tered

-0.67

POSITIVE LOGITS

NP

1.03

NP

0.91

oint

0.77

SS

0.77

ointed

0.73

PP

0.72

emonic

0.72

ublic

0.71

EED

0.71

hemer

0.71

Activations Density 0.006%

the acronym "NP" followed by a single-digit number

references to named entities or categories in a document

No Comments

No Known Activations