INDEX

Explanations

words related to destructive or violent actions involving physical harm

oai_token-act-pair · gpt-3.5-turbo

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

cluding

-0.72

 Hist

-0.66

dds

-0.66

hm

-0.63

history

-0.62

inventoryQuantity

-0.61

CTV

-0.61

since

-0.60

EY

-0.60

 Terms

-0.59

POSITIVE LOGITS

 crept

1.13

 emerged

1.11

 emerges

1.10

 sprang

1.10

 collided

1.09

 suddenly

1.04

 popped

1.03

 intervened

1.02

 approached

1.01

 swoop

0.98

Activations Density 0.394%

words related to destructive or violent actions involving physical harm

No Comments

No Known Activations