INDEX

Explanations

phrases affirming statements or knowledge

oai_token-act-pair · gpt-3.5-turbo

instances of the word "this"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

Ń·

-0.80

»Ĵ

-0.78

ickets

-0.78

pps

-0.77

BuyableInstoreAndOnline

-0.75

ARS

-0.74

veyard

-0.74

å§«

-0.72

oking

-0.71

aughtered

-0.71

POSITIVE LOGITS

 trope

1.09

 phenomenon

1.04

 applies

1.00

 week

0.97

 discrepancy

0.94

 happens

0.91

 tactic

0.88

 pecul

0.88

 possibility

0.88

 behaviour

0.88

Activations Density 0.085%

phrases affirming statements or knowledge

instances of the word "this"

No Comments

No Known Activations