INDEX

Explanations

exaggeratedly positive or negative sentiments

oai_token-act-pair · gpt-3.5-turbo

intensifiers or adverbs expressing strong opinions or feelings

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

antry

-0.81

Previously

-0.68

icipated

-0.68

amide

-0.63

 Altern

-0.63

iem

-0.62

 Previously

-0.61

udic

-0.61

ttes

-0.60

artment

-0.59

POSITIVE LOGITS

 neat

0.90

 nice

0.86

 messed

0.86

 shitty

0.83

 cool

0.83

 pissed

0.81

 crappy

0.81

nice

0.77

 freaking

0.76

bad

0.76

Activations Density 0.072%

exaggeratedly positive or negative sentiments

intensifiers or adverbs expressing strong opinions or feelings

No Comments

No Known Activations