INDEX

Explanations

phrases related to abstract concepts, such as 'the most', 'the biggest', 'the worst'

oai_token-act-pair · gpt-3.5-turbo

repeated references to specific measurements or quantities

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

Allows

-0.88

rade

-0.66

ogie

-0.64

Raid

-0.63

SPONSORED

-0.63

ENCE

-0.63

olo

-0.60

TD

-0.60

 Want

-0.60

ARDIS

-0.59

POSITIVE LOGITS

 earliest

1.04

ses

1.02

 heaviest

0.96

 finest

0.95

 toughest

0.93

 hottest

0.90

 hars

0.88

 strongest

0.87

 greatest

0.86

 hardest

0.85

Activations Density 0.170%

phrases related to abstract concepts, such as 'the most', 'the biggest', 'the worst'

repeated references to specific measurements or quantities

No Comments

No Known Activations