INDEX

Explanations

words related to misinformation or deception

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "mis-" indicating incorrectness or failure

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hetti

-0.76

Â¯Â¯Â¯Â¯

-0.72

jriwal

-0.69

 fixtures

-0.61

INGS

-0.61

eteria

-0.61

 Dynamics

-0.60

 Enhancement

-0.60

 unto

-0.60

 Site

-0.60

POSITIVE LOGITS

cellaneous

1.31

appropri

1.20

pell

1.06

pelled

1.01

beh

0.99

aligned

0.98

informed

0.95

jud

0.94

named

0.92

behavior

0.91

Activations Density 0.014%

words related to misinformation or deception

instances of the prefix "mis-" indicating incorrectness or failure

No Comments

No Known Activations