INDEX

Explanations

words indicating opinions or speculation

oai_token-act-pair · gpt-3.5-turbo

speculative language and expressions of opinion

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 srfAttach

-0.89

oken

-0.72

Interstitial

-0.70

Events

-0.69

 Alone

-0.68

BG

-0.68

CLASSIFIED

-0.66

Done

-0.66

Vs

-0.65

Completed

-0.64

POSITIVE LOGITS

 argue

1.40

 speculated

1.21

 speculate

1.21

 accuse

1.19

 errone

1.17

 contend

1.17

 mistakenly

1.16

 argued

1.16

 criticize

1.15

 theor

1.15

Activations Density 0.200%

words indicating opinions or speculation

speculative language and expressions of opinion

No Comments

No Known Activations