INDEX

Explanations

adjectives related to intensity or depth, particularly negative emotions like concern, sadness, and offense

oai_token-act-pair · gpt-3.5-turbo

intense expressions of concern or distress

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

othy

-0.80

 Millennium

-0.80

amins

-0.77

oras

-0.75

ioch

-0.75

ordan

-0.75

orio

-0.74

iton

-0.70

culosis

-0.69

arters

-0.69

POSITIVE LOGITS

 deeply

0.96

 entrenched

0.95

 ingrained

0.92

 aroused

0.90

 seated

0.87

 rooted

0.86

 enough

0.85

 penetrating

0.85

 immersed

0.83

 embedded

0.83

Activations Density 0.008%

adjectives related to intensity or depth, particularly negative emotions like concern, sadness, and offense

intense expressions of concern or distress

No Comments

No Known Activations