INDEX

Explanations

phrases related to connections or being linked

oai_token-act-pair · gpt-3.5-turbo

instances of the word "connected"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

orically

-0.70

ãĤ¡

-0.67

bra

-0.66

 Leaves

-0.64

ãĤ§

-0.62

YING

-0.61

Swe

-0.60

 Irwin

-0.60

 Lucia

-0.59

VS

-0.59

POSITIVE LOGITS

connected

1.20

 connected

1.19

icut

0.95

Connect

0.93

 connectivity

0.93

 connections

0.91

 Connect

0.90

connect

0.90

 connect

0.89

 dots

0.85

Activations Density 0.011%

phrases related to connections or being linked

instances of the word "connected"

No Comments

No Known Activations