INDEX

Explanations

words related to reporting an event or information

oai_token-act-pair · gpt-3.5-turbo

instances of reporting or feedback in a document

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

ç¥ŀ

-0.77

ophers

-0.69

 ther

-0.60

 booted

-0.60

posted

-0.58

 toast

-0.58

NESS

-0.58

 preserves

-0.57

posts

-0.57

 conferred

-0.57

POSITIVE LOGITS

 Household

0.77

 sightings

0.76

ewitness

0.73

 Circus

0.68

 earthqu

0.67

requency

0.65

 irregularities

0.64

 Fact

0.63

ply

0.62

lehem

0.62

Activations Density 0.231%

words related to reporting an event or information

instances of reporting or feedback in a document

No Comments

No Known Activations