INDEX

Explanations

text related to recaps of events or narratives

oai_token-act-pair · gpt-3.5-turbo

instances of the word "recap" and related terms indicating summary or review content

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

objects

-0.81

 ********************************

-0.75

 Mehran

-0.73

minecraft

-0.73

obyl

-0.69

message

-0.68

Alice

-0.66

worm

-0.66

Initialized

-0.65

comm

-0.65

POSITIVE LOGITS

 recap

2.48

 flex

1.71

 Recap

1.46

 clinch

1.19

ND

1.19

 Patriot

1.10

 rundown

1.10

 Nunes

1.05

flex

1.02

ND

0.97

Activations Density 0.033%

text related to recaps of events or narratives

instances of the word "recap" and related terms indicating summary or review content

No Comments

No Known Activations