INDEX

Explanations

comparative and superlative adjectives related to goodness or benefit

oai_token-act-pair · gpt-3.5-turbo

terms indicating positivity or improvement

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

pter

-0.81

onut

-0.71

dot

-0.69

*/(

-0.68

dit

-0.68

yss

-0.67

ancies

-0.66

abies

-0.65

mares

-0.65

ADRA

-0.65

POSITIVE LOGITS

enough

0.96

 enough

0.93

 suited

0.91

 Enough

0.82

aligned

0.74

 correlated

0.73

 ambassadors

0.70

 safest

0.70

wired

0.70

 aligned

0.67

Activations Density 0.127%

comparative and superlative adjectives related to goodness or benefit

terms indicating positivity or improvement

No Comments

No Known Activations