INDEX

Explanations

quantitative values related to percentages and numerical statistics

oai_token-act-pair · gpt-3.5-turbo

statistical percentages indicating changes or reductions in various contexts

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

ibaba

-0.87

vind

-0.73

ruby

-0.64

agara

-0.64

wait

-0.63

fax

-0.63

achelor

-0.62

etsk

-0.62

abis

-0.61

 âĢº

-0.61

POSITIVE LOGITS

%.

0.77

 mile

0.75

dB

0.72

 thirds

0.67

 margin

0.67

 decreasing

0.66

ciating

0.66

 successive

0.64

 leaps

0.64

 descending

0.63

Activations Density 0.176%

quantitative values related to percentages and numerical statistics

statistical percentages indicating changes or reductions in various contexts

No Comments

No Known Activations