INDEX

Explanations

adjectives related to emotional states, such as worried, concerned, horrified, and grateful

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ilege

-0.81

utra

-0.76

orthy

-0.75

Downloadha

-0.73

cellence

-0.70

interstitial

-0.70

limited

-0.69

istry

-0.69

arin

-0.69

utable

-0.69

POSITIVE LOGITS

 onlook

1.35

 bystanders

1.17

 parents

1.11

 spectators

1.08

 glances

1.06

 villagers

1.05

 shoppers

1.04

 citizens

1.04

 crowds

1.02

 locals

1.01

Activations Density 0.172%

adjectives related to emotional states, such as worried, concerned, horrified, and grateful

No Comments

No Known Activations