INDEX

Explanations

phrases related to hiding or concealing information or events

oai_token-act-pair · gpt-3.5-turbo

phrases related to the concepts of sweeping and being swept, often in a metaphorical context indicating concealment or neglect

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

bia

-0.71

cil

-0.64

rious

-0.63

 Minotaur

-0.62

ISON

-0.62

interrupted

-0.60

named

-0.59

nect

-0.57

 Fraser

-0.57

 Ferry

-0.57

POSITIVE LOGITS

stakes

1.79

 aside

0.90

 away

0.88

apon

0.85

sheets

0.78

soType

0.78

 sweep

0.77

brush

0.76

restling

0.75

 swept

0.75

Activations Density 0.036%

phrases related to hiding or concealing information or events

phrases related to the concepts of sweeping and being swept, often in a metaphorical context indicating concealment or neglect

No Comments

No Known Activations