INDEX

Explanations

uncommonly used words or phrases

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate relative comparisons or contrasts

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

ais

-0.70

visor

-0.69

inus

-0.68

assadors

-0.66

adder

-0.66

ondo

-0.66

frey

-0.65

iseum

-0.65

elin

-0.64

atorium

-0.64

POSITIVE LOGITS

 than

0.99

 anymore

0.95

 altogether

0.92

WARE

0.87

nor

0.86

erous

0.79

than

0.78

 Than

0.77

 bother

0.77

 comprom

0.71

Activations Density 0.158%

uncommonly used words or phrases

phrases that indicate relative comparisons or contrasts

No Comments

No Known Activations