INDEX

Explanations

terms describing a specific concept and then labeling it with a rating

oai_token-act-pair · gpt-3.5-turbo

repetitive phrases or patterns in language, particularly the word "call" followed by various contexts

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

quist

-0.71

idth

-0.68

aban

-0.65

umin

-0.64

estamp

-0.61

etheus

-0.61

sburg

-0.60

 ILCS

-0.59

erial

-0.59

ersen

-0.59

POSITIVE LOGITS

"#

0.71

 bluff

0.71

ovie

0.64

owicz

0.63

geant

0.60

illus

0.60

STDOUT

0.58

sbm

0.58

ones

0.58

Activations Density 0.058%

terms describing a specific concept and then labeling it with a rating

repetitive phrases or patterns in language, particularly the word "call" followed by various contexts

No Comments

No Known Activations