INDEX

Explanations

instances where actions or responses are being described

oai_token-act-pair · gpt-3.5-turbo

instances of the word "react" in various contexts

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

 Highlands

-0.70

ipl

-0.66

 Prin

-0.66

 Belt

-0.66

 Guinness

-0.61

Lord

-0.61

 Membership

-0.61

Pacific

-0.61

Ã

-0.60

Mbps

-0.60

POSITIVE LOGITS

 react

3.82

react

2.53

 reacting

2.38

 reacts

2.34

 reacted

2.30

 React

1.92

 reactions

1.82

 reaction

1.76

 respond

1.62

 reactive

1.53

Activations Density 0.011%

instances where actions or responses are being described

instances of the word "react" in various contexts

No Comments

No Known Activations