INDEX

Explanations

words related to code execution and syntax

oai_token-act-pair · gpt-3.5-turbo

instances of the word "for."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ivil

-0.67

illin

-0.67

beat

-0.66

mare

-0.64

reat

-0.64

âĶ

-0.63

wait

-0.62

nil

-0.62

news

-0.61

Russ

-0.61

POSITIVE LOGITS

 instance

1.19

bidden

1.15

gery

1.15

 example

1.15

 purposes

1.07

geries

1.05

 starters

1.04

ummies

0.90

 debugging

0.89

cing

0.87

Activations Density 0.299%

words related to code execution and syntax

instances of the word "for."

No Comments

No Known Activations