INDEX

Explanations

references to specific TV show titles or channels

oai_token-act-pair · gpt-3.5-turbo

keywords related to television programming or events

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

itect

-0.74

piece

-0.74

ilings

-0.69

orest

-0.69

etter

-0.68

crim

-0.67

ounds

-0.66

blast

-0.65

icons

-0.62

ancers

-0.61

POSITIVE LOGITS

WN

3.10

Else

1.47

Planet

1.25

íķ

1.01

loo

0.88

Sov

0.75

NK

0.74

ESV

0.73

ONSORED

0.69

 Goodbye

0.67

Activations Density 0.011%

references to specific TV show titles or channels

keywords related to television programming or events

No Comments

No Known Activations