INDEX

Explanations

large numerical values, specifically the word "thousand"

oai_token-act-pair · gpt-3.5-turbo

references to the phrase "a thousand."

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

odcast

-0.88

akening

-0.88

livious

-0.85

enture

-0.82

rica

-0.82

NetMessage

-0.81

inion

-0.79

regon

-0.79

untu

-0.78

enhagen

-0.76

POSITIVE LOGITS

oxy

0.81

 snakes

0.72

yen

0.67

Ake

0.65

 injection

0.64

 stripes

0.64

 lions

0.63

 Okin

0.63

 cubic

0.62

 miles

0.62

Activations Density 0.029%

large numerical values, specifically the word "thousand"

references to the phrase "a thousand."

No Comments

No Known Activations