INDEX

Explanations

commands or suggestions to take certain actions

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

©¶æ¥µ

-0.79

HCR

-0.70

folios

-0.69

NetMessage

-0.69

ãĥķ

-0.69

 Estimates

-0.67

ãĥ¼ãĥ³

-0.65

Afee

-0.64

Ts

-0.63

ilib

-0.63

POSITIVE LOGITS

 succeed

0.92

 properly

0.87

 sufficiently

0.87

 slightest

0.85

 succeeds

0.85

 somehow

0.84

 someday

0.79

 correctly

0.79

 fails

0.78

 weren

0.76

Activations Density 4.262%

commands or suggestions to take certain actions

No Comments

No Known Activations