INDEX

Explanations

instances of events or actions being initiated or commenced

oai_token-act-pair · gpt-3.5-turbo

instances of the word "entered" 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 @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Archangel

-0.71

axter

-0.69

pedia

-0.67

roxy

-0.66

killer

-0.64

ector

-0.63

illac

-0.60

 sled

-0.59

WARE

-0.58

hattan

-0.57

POSITIVE LOGITS

prise

1.07

prising

1.06

prises

1.02

tain

0.81

ardless

0.79

tainment

0.76

apore

0.75

igmatic

0.74

èª

0.72

iry

0.71

Activations Density 0.018%

instances of events or actions being initiated or commenced

instances of the word "entered" in various contexts

No Comments

No Known Activations