INDEX

Explanations

phrases related to composition and alignment

oai_token-act-pair · gpt-3.5-turbo

references to elements and compounds

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

-0.74

vier

-0.74

cture

-0.71

 compares

-0.71

rand

-0.71

HUD

-0.71

kef

-0.69

 fooled

-0.69

 followed

-0.68

zed

-0.67

POSITIVE LOGITS

 rest

1.13

 existing

1.11

 broader

0.98

 aforementioned

0.95

 usual

0.90

 wider

0.90

 larger

0.89

 actual

0.88

 pree

0.88

 corresponding

0.84

Activations Density 0.393%

phrases related to composition and alignment

references to elements and compounds

No Comments

No Known Activations