INDEX

Explanations

phrases that indicate conclusions or summaries, often emphasizing a "bottom line."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

roups

-0.77

estern

-0.76

ãĥĺ

-0.76

DOM

-0.74

TIT

-0.74

chlor

-0.72

itialized

-0.71

foundland

-0.71

 UCHIJ

-0.71

lat

-0.70

POSITIVE LOGITS

 boils

0.94

 takeaway

0.82

nings

0.82

 message

0.78

 payoff

0.72

 verdict

0.70

 savings

0.68

 perspective

0.68

 margin

0.67

 economics

0.67

Activations Density 0.007%

phrases that indicate conclusions or summaries, often emphasizing a "bottom line."

No Comments

No Known Activations