INDEX

Explanations

phrases related to acts of compliance or refusal

oai_token-act-pair · gpt-3.5-turbo

instances of the word "be" in various contexts related to permission and conditions

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

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

 sophistication

-0.73

 impossibility

-0.72

 Fail

-0.72

 culmination

-0.70

 Fiction

-0.69

 inconsistency

-0.68

Wid

-0.67

 realization

-0.67

 determination

-0.66

 Disaster

-0.66

POSITIVE LOGITS

 punished

1.17

 interviewed

1.15

 transported

1.13

 deported

1.11

 photographed

1.10

 interrogated

1.10

 hanged

1.08

 escorted

1.08

 treated

1.08

 sacrificed

1.04

Activations Density 0.148%

phrases related to acts of compliance or refusal

instances of the word "be" in various contexts related to permission and conditions

No Comments

No Known Activations