INDEX

Explanations

terms associated with deception or misleading information

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

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L3_Hcat_z_lr1.20e-03_l19.00e-01_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.3.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.02

1:0.02

2:0.10

3:0.11

4:0.01

5:0.03

6:0.07

7:0.18

8:0.06

9:0.04

10:0.08

11:0.21

Negative Logits

 mosqu

-1.19

 todd

-1.03

 srfAttach

-1.03

 enthusi

-1.01

noon

-0.97

FORE

-0.97

ossibility

-0.95

 breeze

-0.92

elsius

-0.92

 upgr

-0.91

POSITIVE LOGITS

 filings

1.03

 omission

0.97

 Hoffman

0.91

antis

0.91

 misrepresent

0.89

 miscar

0.88

 explanations

0.87

nor

0.87

 Sachs

0.87

sites

0.87

Activations Density 0.016%

terms associated with deception or misleading information

No Comments

No Known Activations