INDEX

Explanations

references to objects or concepts described as "ultra"

oai_token-act-pair · gpt-3.5-turbo

references to "ultra" descriptors indicating high performance or quality

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

avan

-0.77

lain

-0.73

 Bleach

-0.73

 Cosponsors

-0.73

 Burnett

-0.70

å£«

-0.69

Interstitial

-0.69

ë

-0.68

emis

-0.67

cific

-0.64

POSITIVE LOGITS

orthodox

0.94

charged

0.90

 ultra

0.81

entimes

0.79

visor

0.77

judicial

0.76

jump

0.75

thodox

0.74

ordinary

0.70

fine

0.69

Activations Density 0.010%

references to objects or concepts described as "ultra"

references to "ultra" descriptors indicating high performance or quality

No Comments

No Known Activations