INDEX

Explanations

phrases related to political events and key officials

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

zsche

-0.83

istic

-0.78

resp

-0.71

istically

-0.70

 commercially

-0.69

company

-0.67

 traffickers

-0.67

ties

-0.65

ãĤ¨ãĥ«

-0.64

 traff

-0.60

POSITIVE LOGITS

 hearings

1.15

woman

1.14

 Speaker

1.08

 adjourn

1.04

 Appropriations

1.03

 subcommittee

1.01

 unanimously

1.00

 Votes

1.00

 Subcommittee

1.00

 Judiciary

1.00

Activations Density 3.395%

phrases related to political events and key officials

No Comments

No Known Activations