INDEX

Explanations

phrases related to declining to comment

oai_token-act-pair · gpt-3.5-turbo

mentions of refusal or declines to provide information

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

abiding

-0.92

soDeliveryDate

-0.65

ersed

-0.61

pires

-0.61

 rotten

-0.60

oplan

-0.60

grass

-0.60

loving

-0.59

righteous

-0.58

RAW

-0.58

POSITIVE LOGITS

 disclose

1.16

 specify

1.14

 discuss

1.12

 speculate

1.10

 comment

1.09

 endorse

1.09

 identify

1.06

 quantify

1.06

 clarify

1.02

 characterize

1.00

Activations Density 0.065%

phrases related to declining to comment

mentions of refusal or declines to provide information

No Comments

No Known Activations