INDEX

Explanations

questions or requests for information directed towards a person or entity

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

Ĥ¬

-0.78

 Scouting

-0.67

âĶĢâĶĢ

-0.66

EStreamFrame

-0.66

luaj

-0.66

cutting

-0.64

lim

-0.64

Lago

-0.62

absor

-0.61

 enactment

-0.61

POSITIVE LOGITS

 questions

1.29

 rhet

1.27

naires

1.14

 answered

1.09

 probing

1.09

 Questions

1.06

naire

0.98

 question

0.96

 asked

0.96

 politely

0.93

Activations Density 1.324%

questions or requests for information directed towards a person or entity

No Comments

No Known Activations