INDEX

Explanations

verbs in base form after the word 'can'

oai_token-act-pair · gpt-3.5-turbo

negations or phrases expressing inability or restrictions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

 Reck

-0.65

Wol

-0.64

 intending

-0.63

Cly

-0.60

 Notting

-0.59

Kub

-0.59

 Respons

-0.58

 guiActiveUnfocused

-0.56

 Fell

-0.55

Oz

-0.55

POSITIVE LOGITS

estine

0.86

atche

0.86

adian

0.78

 safely

0.78

osta

0.77

rouse

0.76

apesh

0.74

 muster

0.74

atell

0.74

 withstand

0.74

Activations Density 0.745%

verbs in base form after the word 'can'

negations or phrases expressing inability or restrictions

No Comments

No Known Activations