INDEX

Explanations

words related to seizing control or leadership

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

Statement

-0.72

cohol

-0.71

spir

-0.70

gru

-0.69

 solicit

-0.68

 graduates

-0.67

ileaks

-0.67

study

-0.66

 accompan

-0.66

IDER

-0.64

POSITIVE LOGITS

 territory

1.06

 foothold

1.04

 possession

0.96

ignty

0.94

 vacated

0.94

 supremacy

0.94

 reins

0.92

 rights

0.91

 control

0.90

 custody

0.90

Activations Density 1.643%

words related to seizing control or leadership

No Comments

No Known Activations