INDEX

Explanations

terms related to entry-level positions or features

oai_token-act-pair · gpt-3.5-turbo

instances of the word "entry" followed by numeric levels

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ecause

-0.72

icz

-0.70

rica

-0.70

rior

-0.65

enburg

-0.64

irit

-0.61

 Band

-0.61

uku

-0.61

nder

-0.60

amily

-0.60

POSITIVE LOGITS

prising

1.03

prises

0.94

prise

0.93

tain

0.92

 into

0.86

 INTO

0.84

tainment

0.84

tained

0.84

way

0.82

ways

0.80

Activations Density 0.032%

terms related to entry-level positions or features

instances of the word "entry" followed by numeric levels

No Comments

No Known Activations