INDEX

Explanations

terms related to specific organizations and possibly names of people

oai_token-act-pair · gpt-3.5-turbo

specific nouns and terms that are related to various subjects, indicating a focus on names and labels

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

å£«

-0.80

swer

-0.65

alogue

-0.62

âĸ¬

-0.60

 corrid

-0.57

 recognised

-0.57

NX

-0.57

ilater

-0.57

 Audit

-0.56

aroo

-0.56

POSITIVE LOGITS

 lymph

0.66

minecraft

0.63

 boil

0.58

 oils

0.56

aneous

0.56

ï¸ı

0.56

rious

0.55

 bodily

0.55

ngth

0.55

nir

0.55

Activations Density 1.268%

terms related to specific organizations and possibly names of people

specific nouns and terms that are related to various subjects, indicating a focus on names and labels

No Comments

No Known Activations