INDEX

Explanations

actions or steps that can be performed

oai_token-act-pair · gpt-3.5-turbo

phrases indicating actions or outcomes related to instructions or controls

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

 Mund

-0.88

mb

-0.79

GEN

-0.78

marg

-0.78

-0.77

 Sylv

-0.76

®

-0.75

 Gray

-0.75

-0.74

POSITIVE LOGITS

ake

1.13

akes

1.11

 Take

1.10

Take

1.06

 Takes

1.05

take

1.05

 take

1.02

 retake

0.96

 taken

0.93

 takes

0.91

Activations Density 0.321%

actions or steps that can be performed

phrases indicating actions or outcomes related to instructions or controls

No Comments

No Known Activations