INDEX

Explanations

instances where items are physically connected or attached to each other

oai_token-act-pair · gpt-3.5-turbo

instances of the word "attached."

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

$$$$

-0.71

Cola

-0.69

java

-0.66

 Kardashian

-0.65

uddin

-0.64

Bos

-0.64

å£«

-0.63

Martin

-0.63

inventoryQuantity

-0.63

externalActionCode

-0.62

POSITIVE LOGITS

 thereto

0.94

 detach

0.92

 attach

0.89

 attached

0.86

ments

0.86

nect

0.77

 comr

0.77

rity

0.76

 securely

0.76

pole

0.75

Activations Density 0.020%

instances where items are physically connected or attached to each other

instances of the word "attached."

No Comments

No Known Activations