INDEX

Explanations

references to a specific place or word, "rat"

oai_token-act-pair · gpt-3.5-turbo

references to a specific group or characterized entities referred to as "rat" in a context suggestive of conflict or terrorism

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

INTER

-0.65

 Mandela

-0.64

conservancy

-0.64

 convol

-0.64

 Lauder

-0.63

ACP

-0.62

 Cyprus

-0.61

 scarce

-0.61

OHN

-0.60

ERC

-0.60

POSITIVE LOGITS

chet

1.15

itude

0.97

isen

0.96

hea

0.92

oxin

0.91

igraph

0.90

rat

0.89

rollers

0.88

hetic

0.88

iger

0.87

Activations Density 0.010%

references to a specific place or word, "rat"

references to a specific group or characterized entities referred to as "rat" in a context suggestive of conflict or terrorism

No Comments

No Known Activations