INDEX

Explanations

phrases related to news and events

oai_token-act-pair · gpt-3.5-turbo

phrases that reference specific details or attributes of subjects

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

Layer

-0.71

thood

-0.70

ĸļ

-0.70

verbs

-0.67

 Ships

-0.64

shell

-0.63

 countered

-0.63

linux

-0.63

lite

-0.63

Ľ

-0.63

POSITIVE LOGITS

 incident

1.16

 discrepancy

1.12

 ordeal

1.10

 latest

1.10

 situation

1.09

 announcement

1.06

 outcome

0.99

 findings

0.98

 plight

0.95

 sudden

0.94

Activations Density 0.612%

phrases related to news and events

phrases that reference specific details or attributes of subjects

No Comments

No Known Activations