INDEX

Explanations

large numerical values or quantifiers

oai_token-act-pair · gpt-3.5-turbo

numerical data related to statistics and percentages

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

Agent

-0.64

 Fraz

-0.60

 Captain

-0.59

Job

-0.59

 Clash

-0.58

 Constantin

-0.58

 Coch

-0.58

 Faust

-0.58

 Hawth

-0.58

 Faul

-0.58

POSITIVE LOGITS

ÃĹ

0.84

 consecutive

0.81

Â½

0.81

%-

0.80

âĢ²

0.78

kHz

0.77

Â¢

0.76

itialized

0.74

yen

0.74

Activations Density 0.196%

large numerical values or quantifiers

numerical data related to statistics and percentages

No Comments

No Known Activations