INDEX

Explanations

The main thing this neuron does is find keywords related to long-term aspects of various topics

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

illard

-0.78

IRO

-0.77

ICAN

-0.74

leck

-0.72

ILLE

-0.71

 Compass

-0.67

ILA

-0.64

unin

-0.61

CB

-0.61

ONSORED

-0.60

POSITIVE LOGITS

itud

1.27

itudinal

1.17

sword

1.17

lasting

1.10

term

1.06

 term

1.04

sighted

1.01

 lasting

1.01

shore

1.01

 overdue

1.00

Activations Density 2.601%

The main thing this neuron does is find keywords related to long-term aspects of various topics

No Comments

No Known Activations