INDEX

Explanations

negations or obstacles

oai_token-act-pair · gpt-3.5-turbo

expressions related to obstacles or difficulties

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 fundamentally

-0.78

 radically

-0.66

matically

-0.66

 Timeline

-0.66

ocracy

-0.65

 historically

-0.62

 inherently

-0.61

ually

-0.61

 Closed

-0.60

 basically

-0.60

POSITIVE LOGITS

 inconvenience

1.11

 consolation

1.07

 avail

0.97

 assistance

0.94

 inconven

0.89

 distraction

0.88

 Advantage

0.85

 welcome

0.85

 annoyance

0.83

 encouragement

0.82

Activations Density 0.483%

negations or obstacles

expressions related to obstacles or difficulties

No Comments

No Known Activations