Samuel Recht

@sam.re

219 Following 1.1k Followers

Cognitive scientist @ Oxford. Interested in attention, metacognition and curiosity during learning.

https://sam.re

@joannabagniewska.com Oct 12, 2026, 3:01 PMReposted by @sam.re

How much money do PhD candidates get across Europe?
It really depends.

Yes - taxes, purchasing power, benefits, etc., all come into play, but the differences are still staggering.

#academicsky
🧪🎓
www.nature.com/articles/d41...

A graph of remuneration (x-axis) for PhD candidates in different countries (y-axis). The Employment-based ones are in orange, the scholarship-based ones in blue.

@meganakpeters.bsky.social Oct 10, 2026, 12:41 PMReposted by @sam.re

Thank you so much to @newscientist.com Live for the opportunty to talk about AI consciousness (& how we really, really don't have a good way of testing for it)!

Headlines & scientists saying "OMG it might be conscious, 5% chance" are sensationalism BS. Based on what? Vibes? Not good enough for me.

@erictopol.bsky.social Oct 8, 2026, 6:10 PMReposted by @sam.re

New @science.org
A breakthrough in understanding why do we get fever and an inflammation response to vaccines.
The gut microbiome chronic production of flagellin
Paves the way to potentially prevent these acute phase side-effects in the future
www.science.org/doi/10.1126/...

@gaiamolinaro.bsky.social Oct 8, 2026, 8:01 PMReposted by @sam.re

Our reward function compression work is now out in Nature Communications! 🎉 In it, we present and empirically test a computational framework that formalizes the trade-offs between goal flexibility and learning efficiency. @annecollins.bsky.social @natureportfolio.nature.com doi.org/10.1038/s414...

Conceptual model of goal-directed reinforcement learning with reward
function compression. In our proposed conceptual model, we break down goald-dependent learning into several parts. Combined with other learning components, an RL learning algorithm informs value estimates W on a given trial t, which are entered into a policy. The agent selects an action a through a joint policy, to which the environment responds by outputting outcome information o and a new state s. A working memory (WM) component organizes and maintains task-relevant information (here, broadly intended and dubbed the task information), used to support the reinforcement learning (RL) process. a Initially, value attribution to o is operated by WM,which maintains the current goal (g) and supports its matching to the experienced outcome to generate a reward r used by the RL agent. WM also holds a trace of recently experienced goals. b If goals share enough similar features,
a compressed reward function can be created, which replaces active goal maintenance and can, over time, be transferred to long-term memory storage. Eventually, the compressed reward function can be directly employed to assign value to experienced outcomes, sparing resources for other WM tasks.

@sam.re Oct 9, 2026, 2:18 PM

very interesting, thank you!

@botoscsabi.bsky.social Oct 8, 2026, 11:44 AMReposted by @sam.re

[Hugging Face release] We’re releasing the dataset behind Reason to Play! I’m incredibly proud of our team for reaching this milestone.

@science.org Oct 4, 2026, 11:13 PMReposted by @sam.re

A recent analysis in Science of more than 116,000 bird songs has revealed that the astonishing variety of passerine birdsong is built from just eight basic acoustic motifs, like trills, whistles, and harmonic notes.

Learn more on #WorldAnimalDay: https://scim.ag/3SziyeL

A phylogenetic distribution and trait correlates of acoustic motifs in passerine birds. Hedline reads: The global biogeography of passerine songs

@smfleming.bsky.social Oct 5, 2026, 9:56 AMReposted by @sam.re

Excited to be launching our new PhD program in Minds and Machines at UCL with co-Directors Tali Sharot and @nadinedijkstra.bsky.social

To apply, find out more here: www.ucl.ac.uk/brain-scienc...

Our launch event is on Nov 10th, with a cracking lineup of speakers: luma.com/cikzcth0

@bayslab.org Oct 5, 2026, 11:28 AMReposted by @sam.re

Do the contents of visual #WorkingMemory deteriorate more quickly when more items are being remembered? Seems a straightforward question but we had to bring together 13 studies and >120,000 trials to get definitive answers... doi.org/10.1037/rev0...

@ivntmc.bsky.social

@samgilbert.bsky.social Oct 5, 2026, 3:27 PMReposted by @sam.re

Just published - what happens in the brain when people offload memories into external reminders? Using fMRI decoding, we found that the corresponding neural trace fades until it becomes statistically absent.

www.nature.com/articles/s41...

🧵...

@natrevpsychol.nature.com Oct 5, 2026, 5:53 PMReposted by @sam.re

Science-informed opinion on generative AI for young children

Comment by Kathy Hirsh-Pasek (@kathyhirshpasek.bsky.social) & Roberta Michnick Golinkoff

go.nature.com/4dkGhql

@norabradford.bsky.social Oct 5, 2026, 4:46 PMReposted by @sam.re

love to see metacognition getting public attention:) www.thetransmitter.org/decision-mak...

@sabomelinda.bsky.social Oct 5, 2026, 3:49 PMReposted by @sam.re

2003 saw the birth of the retro-cue task. 23 years later, we adapt this logic to LTM and ask how attention shapes LTM retrieval. Check out our paper in Communications Psychology: www.nature.com/articles/s44...
w/ @william-nm.bsky.social, Kia, and lab alumni Nahid and @norarouast.bsky.social 1/n

@earlkmiller.bsky.social Oct 4, 2026, 1:02 PMReposted by @sam.re

When you focus on one item in working memory, beta rhythms may help “park” the other in a protected, set-aside state.

Prioritization in Working Memory Reduces Interference via Beta Band-Linked Stimulus Transformations
doi.org/10.1523/JNEU...
#neuroscience

@benhayden.bsky.social Oct 2, 2026, 7:52 PMReposted by @sam.re

New paper from the Neurosurgery research team at BCM! 🧵

"A population code for semantics in human hippocampus"

led by Melissa Franch. We recorded hippocampus single neurons while people listened to podcasts.

www.nature.com/articles/s41...

@cstrauch.bsky.social Oct 2, 2026, 5:45 AMReposted by @sam.re

How much of a risk are LLM-generated bots for online behavioral experiments?
We present results of an entirely natural language prompt-generated agent, capable of absolving any of the 26 experiments we tested it on. Most tasks are passed with data that can pass as human osf.io/preprints/ps...
(1/)

@richardhuskey.bsky.social Oct 2, 2026, 10:07 PMReposted by @sam.re

🧵In April, I built an AI agent that could complete the Attention Network Task. The agent was task-specific and costly. I concluded that online behavioral experiments were safe from agentic infiltration. Six months later, my view has changed. Our new preprint explains why osf.io/preprints/ps...

@kobedesender.bsky.social Sep 29, 2026, 8:17 AMReposted by @sam.re

"Trial-by-trial fluctuations in decision criterion shape confidence"! Now out in @nathumbehav.nature.com led by Robin Vloeberghs, Lara Navarette, @anne-urai.bsky.social and me.
PDF: desenderlab.com/wp-content/u... Fluctuating Thread ↓↓↓

@jayvanbavel.bsky.social Sep 24, 2026, 4:29 PMReposted by @sam.re

Cornell produced new report on higher education noting that companies have a “jarring” skepticism of higher ed thanks to a new generation of degree holders who lack key skills like handling uncertainty, compromising, and absorbing feedback. Faculty also report that students lack these skills.

@deevybee.bsky.social Sep 24, 2026, 4:14 PMReposted by @sam.re

New (sorry, rather long) blogpost on Reproducibility Packages, garden of forking paths, and importance of checking validity of datasets and analyses.
deevybee.blogspot.com/2026/09/trus...
This expands on my PubPeer comments on Packalen & Bhattacharya (2020) article. pubpeer.com/publications...

Older posts