INDEX

Explanations

full names or proper nouns, possibly related to news or events occurring in a specific context

oai_token-act-pair · gpt-3.5-turbo

occurrences of a specific symbol or punctuation mark

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 obser

-0.91

 conception

-0.76

awaru

-0.75

 puff

-0.75

 imagination

-0.74

 ende

-0.73

 imperson

-0.72

 downed

-0.72

 unconscious

-0.71

 halluc

-0.70

POSITIVE LOGITS

Â¯

1.03

ï¸ı

0.85

âĢł

0.85

said

0.84

tab

0.81

tra

0.81

tre

0.81

Â£

0.79

âĪ

0.77

Â°

0.77

Activations Density 0.269%

full names or proper nouns, possibly related to news or events occurring in a specific context

occurrences of a specific symbol or punctuation mark

No Comments

No Known Activations