INDEX

Explanations

words related to warnings or alerts, especially those related to potential negative consequences

oai_token-act-pair · gpt-3.5-turbo

references to "triggers" in various contexts

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ensable

-0.77

hemat

-0.75

apest

-0.74

nian

-0.73

apolis

-0.73

 Cutler

-0.70

cott

-0.70

ately

-0.69

ijk

-0.66

esan

-0.66

POSITIVE LOGITS

trigger

0.96

 triggering

0.96

 warnings

0.93

 triggers

0.90

 trigger

0.81

 Trigger

0.76

 triggered

0.74

witz

0.73

 Warn

0.73

 alerts

0.73

Activations Density 0.050%

words related to warnings or alerts, especially those related to potential negative consequences

references to "triggers" in various contexts

No Comments

No Known Activations