INDEX

Explanations

numeric units such as distances or time durations

oai_token-act-pair · gpt-3.5-turbo

references to the previous year, specifically the term "last year."

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

 textual

-0.86

 encl

-0.72

 cumbers

-0.71

 zipper

-0.71

 showc

-0.70

 chains

-0.68

ogly

-0.68

hod

-0.67

 grids

-0.67

cro

-0.64

POSITIVE LOGITS

long

0.96

book

0.89

ago

0.76

âĢ²

0.76

iannopoulos

0.75

nings

0.74

 olds

0.74

 anniversary

0.73

iversary

0.72

eve

0.70

Activations Density 0.080%

numeric units such as distances or time durations

references to the previous year, specifically the term "last year."

No Comments

No Known Activations