INDEX

Explanations

mentions of actions that are emotionally charged

oai_token-act-pair · gpt-3.5-turbo

emotional expressions and requests for interaction or responses

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

 interfaces

-0.81

 corridors

-0.80

 bills

-0.79

 branches

-0.79

 teachings

-0.77

 channels

-0.77

 practices

-0.76

 establishments

-0.76

 timelines

-0.76

 avenues

-0.76

POSITIVE LOGITS

 somewhere

0.96

 apiece

0.82

ful

0.77

breaker

0.76

robe

0.76

 worthy

0.76

 breaker

0.74

pair

0.74

itamin

0.72

naire

0.72

Activations Density 0.381%

mentions of actions that are emotionally charged

emotional expressions and requests for interaction or responses

No Comments

No Known Activations