INDEX

Explanations

questions directed to someone, possibly as part of a conversation or interview

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "ask" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

undo

-0.69

lim

-0.64

pite

-0.64

zinski

-0.62

anking

-0.60

cutting

-0.59

ovie

-0.59

rongh

-0.58

swing

-0.58

cyclop

-0.57

POSITIVE LOGITS

 questions

1.03

 rhet

1.02

wered

1.00

 probing

0.95

 answ

0.90

 Questions

0.84

 answered

0.83

 permission

0.81

 politely

0.81

naires

0.81

Activations Density 0.050%

questions directed to someone, possibly as part of a conversation or interview

occurrences of the word "ask" in various contexts

No Comments

No Known Activations