INDEX

Explanations

certain keywords or phrases that signal a timing or prompt for action

oai_token-act-pair · gpt-3.5-turbo

references to cues and signals 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 @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 ILCS

-0.80

icum

-0.77

abeth

-0.73

ibaba

-0.69

 Armenian

-0.69

ategory

-0.69

adders

-0.68

oslav

-0.67

ended

-0.66

ends

-0.65

POSITIVE LOGITS

cue

1.01

llor

0.99

ball

0.92

balls

0.89

Cue

0.86

ption

0.82

eer

0.82

 cues

0.81

eering

0.80

BOOK

0.78

Activations Density 0.021%

certain keywords or phrases that signal a timing or prompt for action

references to cues and signals in various contexts

No Comments

No Known Activations