INDEX

Explanations

terms related to news reporting and events, particularly statements indicating official action or responses

oai_token-act-pair · gpt-3.5-turbo

references to requests for comments or responses in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 abandon

-0.67

 smile

-0.62

 abandoning

-0.61

 life

-0.61

 sort

-0.61

 lick

-0.61

 watering

-0.61

 herself

-0.59

ython

-0.59

 retreat

-0.59

POSITIVE LOGITS

Meanwhile

1.14

Also

1.12

However

1.09

Earlier

1.07

Previously

1.06

Additionally

1.05

According

1.03

Sources

1.02

Asked

0.98

Additional

0.98

Activations Density 0.640%

terms related to news reporting and events, particularly statements indicating official action or responses

references to requests for comments or responses in various contexts

No Comments

No Known Activations