INDEX

Explanations

instances of the word "false"

oai_token-act-pair · gpt-3.5-turbo

phrases related to false claims or deception

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

hens

-1.00

hetti

-0.94

ODY

-0.84

imen

-0.82

scene

-0.80

mun

-0.79

 guiActiveUnfocused

-0.79

APTER

-0.76

arya

-0.74

ILE

-0.73

POSITIVE LOGITS

 accuser

0.94

 false

0.93

 guiActiveUn

0.92

 unfocusedRange

0.88

 positives

0.87

 misrepresent

0.84

 dich

0.83

 falsely

0.80

 guiIcon

0.78

false

0.77

Activations Density 0.014%

instances of the word "false"

phrases related to false claims or deception

No Comments

No Known Activations