INDEX

Explanations

terms related to hierarchical classification levels or types

oai_token-act-pair · gpt-3.5-turbo

classifications of legal offenses or classes of individuals

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

OPLE

-0.73

 stump

-0.69

wit

-0.68

 Bakr

-0.66

 ende

-0.62

issance

-0.61

ternity

-0.58

 Senegal

-0.58

 Scotia

-0.58

 Sussex

-0.58

POSITIVE LOGITS

ifications

1.33

ifiers

1.09

ifying

1.07

ifies

1.05

ifier

1.04

mates

1.00

ifiable

0.97

mate

0.96

ribing

0.92

ification

0.90

Activations Density 0.021%

terms related to hierarchical classification levels or types

classifications of legal offenses or classes of individuals

No Comments

No Known Activations