INDEX

Explanations

references to a specific organization or company with the acronym "WA"

oai_token-act-pair · gpt-3.5-turbo

references to labor unions or union-related activities

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

nir

-0.82

lio

-0.74

named

-0.74

tie

-0.73

sets

-0.73

nam

-0.73

names

-0.72

meyer

-0.71

raph

-0.71

oslav

-0.69

POSITIVE LOGITS

VE

1.28

BILITY

1.16

HA

1.14

ILA

1.14

ZE

1.14

WA

1.11

HE

1.08

UGH

1.07

ZA

1.07

HAHA

1.06

Activations Density 0.037%

references to a specific organization or company with the acronym "WA"

references to labor unions or union-related activities

No Comments

No Known Activations