INDEX

Explanations

adjectives related to the evaluation of concepts or things

oai_token-act-pair · gpt-3.5-turbo

phrases related to opinions or assessments of societal and cultural topics

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

igent

-0.77

 conclud

-0.74

steps

-0.70

ktop

-0.70

Materials

-0.69

 underscores

-0.69

 reiterate

-0.68

cellaneous

-0.68

edIn

-0.68

»Ĵ

-0.67

POSITIVE LOGITS

 ruining

1.16

 gonna

1.10

 racist

1.09

 sexist

1.04

 extinct

1.04

 crap

1.04

nt

1.02

 unbeat

1.02

BAD

1.00

 horrible

1.00

Activations Density 0.406%

adjectives related to the evaluation of concepts or things

phrases related to opinions or assessments of societal and cultural topics

No Comments

No Known Activations