INDEX

Explanations

references to various types of awards

oai_token-act-pair · gpt-3.5-turbo

mentions of awards and recognition

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

 paralle

-0.74

versions

-0.73

AppData

-0.72

nels

-0.70

 smokes

-0.65

verse

-0.64

 resisted

-0.64

 Mono

-0.63

Weak

-0.63

nesses

-0.62

POSITIVE LOGITS

 award

3.96

 Award

2.74

 awards

2.66

 Awards

2.08

 awarding

2.08

 awarded

1.92

 Emmy

1.57

 accol

1.54

 prize

1.52

 Grammy

1.34

Activations Density 0.008%

references to various types of awards

mentions of awards and recognition

No Comments

No Known Activations