INDEX

Explanations

phrases related to accusations or responsibility

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

 horm

-0.71

 acquaintances

-0.68

types

-0.67

onyms

-0.67

 guiActiveUnfocused

-0.67

ibl

-0.66

fw

-0.65

erm

-0.64

 await

-0.64

ieties

-0.62

POSITIVE LOGITS

 overseeing

1.10

 shaping

0.87

 steering

0.86

 organising

0.85

 orchestr

0.84

 brunt

0.84

 superv

0.80

 organizing

0.78

 coordinating

0.78

 guiding

0.77

Activations Density 5.720%

phrases related to accusations or responsibility

No Comments

No Known Activations