INDEX

Explanations

years expressed as four-digit numbers

oai_token-act-pair · gpt-3.5-turbo

mentions of the year 2012

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

anooga

-0.86

imore

-0.80

avascript

-0.80

ï¸

-0.80

merce

-0.79

itional

-0.77

onite

-0.77

ournals

-0.77

itionally

-0.76

phy

-0.75

POSITIVE LOGITS

å¹

0.87

nd

0.84

 onwards

0.75

0.72

 Reincarn

0.71

-'

0.70

ploy

0.68

0.67

cens

0.67

0.66

Activations Density 0.021%

years expressed as four-digit numbers

mentions of the year 2012

No Comments

No Known Activations