INDEX

Explanations

sequences that look like dates

oai_token-act-pair · gpt-3.5-turbo

lists or collections of numbered releases, particularly those associated with media releases over various years

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

oshenko

-0.67

arium

-0.62

SR

-0.62

renheit

-0.59

Mn

-0.58

Lago

-0.58

ERAL

-0.57

oidal

-0.56

emonic

-0.55

iator

-0.54

POSITIVE LOGITS

comings

0.69

ãĥ¼ãĥĨãĤ£

0.65

 conclud

0.62

church

0.57

hill

0.57

gio

0.56

 belonged

0.55

å¤©

0.55

ãģ£

0.55

ahime

0.54

Activations Density 0.101%

sequences that look like dates

lists or collections of numbered releases, particularly those associated with media releases over various years

No Comments

No Known Activations