INDEX

Explanations

words related to directions, specifically focusing on the sides (left and right)

oai_token-act-pair · gpt-3.5-turbo

references to the left and right orientations or positions

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

ãĥīãĥ©

-0.83

oulos

-0.82

BuyableInstoreAndOnline

-0.79

doms

-0.73

onis

-0.72

ITED

-0.72

è¦ļéĨĴ

-0.72

ifi

-0.71

zan

-0.71

aters

-0.70

POSITIVE LOGITS

 side

1.24

 flank

1.13

 wing

1.10

 hand

1.09

most

1.09

leg

1.05

 hemisphere

1.04

 forearm

1.02

 thigh

1.02

 edge

0.98

Activations Density 0.046%

words related to directions, specifically focusing on the sides (left and right)

references to the left and right orientations or positions

No Comments

No Known Activations