INDEX

Explanations

activations related to specific phrases or patterns that may not easily lend themselves to interpretation without additional context

oai_token-act-pair · gpt-3.5-turbo

repeated references to the abbreviation "BL" associated with various values

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

enegger

-0.90

framework

-0.83

geist

-0.80

aeda

-0.74

gerald

-0.74

mble

-0.70

oteric

-0.70

tainment

-0.70

eers

-0.66

 Authorization

-0.65

POSITIVE LOGITS

OOD

0.99

ACK

0.97

anca

0.88

OW

0.87

adder

0.87

ighting

0.86

OCK

0.85

umenthal

0.85

pedia

0.84

obl

0.79

Activations Density 0.004%

activations related to specific phrases or patterns that may not easily lend themselves to interpretation without additional context

repeated references to the abbreviation "BL" associated with various values

No Comments

No Known Activations