INDEX

Explanations

names or terms ending with 'ho'

oai_token-act-pair · gpt-3.5-turbo

repeated instances of the word "ho."

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

ãĥ¡

-0.82

ividual

-0.77

iosyncr

-0.77

AAF

-0.66

rations

-0.65

UCT

-0.64

Dialogue

-0.63

ifications

-0.62

ication

-0.62

downs

-0.62

POSITIVE LOGITS

ppy

0.94

efully

0.92

ppo

0.91

vernment

0.90

ffer

0.90

ancock

0.86

ven

0.84

pping

0.83

arding

0.83

pper

0.82

Activations Density 0.015%

names or terms ending with 'ho'

repeated instances of the word "ho."

No Comments

No Known Activations