INDEX

Explanations

proper nouns that are likely related to news articles or other journalistic content

oai_token-act-pair · gpt-3.5-turbo

punctuation or formatting used in references or annotations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 Squadron

-0.67

swer

-0.67

 endeavour

-0.64

 consideration

-0.61

 Deity

-0.61

 Honour

-0.59

lc

-0.58

gery

-0.58

dq

-0.57

qv

-0.57

POSITIVE LOGITS

WASHINGTON

0.92

Washington

0.83

ccording

0.79

NEW

0.74

SAN

0.73

GREEN

0.68

ente

0.67

Dallas

0.66

VERTISEMENT

0.66

Loading

0.66

Activations Density 0.188%

proper nouns that are likely related to news articles or other journalistic content

punctuation or formatting used in references or annotations

No Comments

No Known Activations