INDEX

Explanations

words related to news

oai_token-act-pair · gpt-3.5-turbo

mentions of news and media

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

inho

-0.70

 Defin

-0.69

 Alto

-0.65

 Stark

-0.65

 blush

-0.65

 faint

-0.65

Lak

-0.63

 Stupid

-0.63

 Hamm

-0.62

 Pearce

-0.61

POSITIVE LOGITS

room

1.34

rooms

1.33

worthy

1.28

worthiness

1.24

feed

1.22

reader

1.20

peak

1.19

print

1.15

stand

1.13

agents

1.12

Activations Density 0.050%

words related to news

mentions of news and media

No Comments

No Known Activations