acronym endings like HSUS and SDS

What this neuron does

This neuron fires on the last letters of capital-letter abbreviations, such as the 'US' in HSUS or the 'S' in SDS. It also goes strongly negative on weekday names in email date headers. Turning it up did not change the model's writing in a clear way.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

1. #1 peak

   beagle saved from a Michigan research lab in 2019 after an HSUS undercover investigation at Charles River Laboratories, is now the namesake for important bipartisan legislation

2. #2 peak

   very strong community of animal welfare advocates.⏎Molly Tamulevich, the HSUS⏎H.B. 4482 requires Michigan labs that use dogs and

3. #3 peak

   processes related to WTO Agreements such as agriculture, sanitary and phytosanitary measures (SPS), technical barriers to trade (TBT), trade-related aspects of intellectual property rights

4. #4 peak

   . Temperature range⏎4. Temperature display accuracy⏎5. Temperature stability⏎6. Temperature uniformity⏎7. Heating time⏎8. Internal material⏎SUS304 stainless steel plate, thickness 1.2

5. #5 peak

   wireless (not all warnings sent to the ship reached the bridge, and a nearby ship, the operator abed, missed Titanic's SOS). Very recently, from recovered wreckage, "Popular Science" claimed the hull was

6. #6 peak

   the entire tree until you see that there is little chance that the tree will<|endoftext|>Safety Data Sheet Information⏎Employers must ensure that the Safety Data Sheet (SDSs) are readily accessible to employees for all hazardous chemicals in their workplace. This

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

1. #1 peak

   chemistryFri, 3 Sep 2010, 12:51:40 EDT⏎- Martian rock arrangement not alien handiworkWed, 7 Jan 2009, 13:

2. #2 peak

   . // in Elliot Hildner / University of Bremen⏎The main crack<|endoftext|>analysis of algoritms⏎robert.kern at gmail.com⏎Fri Sep 10 00:10:22 CEST 2

3. #3 peak

   4 + 314 = 512<|endoftext|>[Haskell-beginners] Question about data structures⏎daniel.is.fischer at web.de⏎Wed Nov

4. #4 peak

   have been clear enough. ------------------ Come on in !! Drinks are on the house in the Big Moose Saloon !!⏎Udayan Naik<BR>Sun Certified Programmer for the Java <|endoftext|>Australia takes about 13 00

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.