INDEX

Explanations

phrases related to actions or descriptions followed by a hyphen and a numerical rating

oai_token-act-pair · gpt-3.5-turbo

negative sentiments or expressions of dissatisfaction

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.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

 Fiscal

-0.90

 Burr

-0.77

 indemn

-0.72

 fiscal

-0.70

AVG

-0.70

 Qualcomm

-0.67

 Burnett

-0.67

 Pompe

-0.67

 Dickinson

-0.67

 dispute

-0.66

POSITIVE LOGITS

sized

1.31

sama

1.21

themed

1.20

dimensional

1.13

conscious

1.11

prototype

1.10

san

1.09

loving

1.08

based

1.06

turned

1.04

Activations Density 0.275%

phrases related to actions or descriptions followed by a hyphen and a numerical rating

negative sentiments or expressions of dissatisfaction

No Comments

No Known Activations