INDEX

Explanations

mentions of government-related terms or processes

oai_token-act-pair · gpt-3.5-turbo

terms related to 'ment' suffixes, indicating various forms of states or processes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

²¾

-0.70

 bite

-0.69

Topics

-0.63

iPhone

-0.63

çĦ

-0.63

pins

-0.63

JB

-0.62

INO

-0.62

soDeliveryDate

-0.60

WT

-0.60

POSITIVE LOGITS

ment

1.54

ments

0.99

ation

0.94

rade

0.94

ary

0.90

ally

0.89

naire

0.86

ement

0.86

ially

0.84

aler

0.82

Activations Density 0.012%

mentions of government-related terms or processes

terms related to 'ment' suffixes, indicating various forms of states or processes

No Comments

No Known Activations