Chris Cooper

Delighted to share that my article 'Predicting the CEFR level of English listening texts with machine learning methods' was just published in Research Methods in Applied Linguistics. Free access for 50 days: sciencedirect.com/science/articl… Thank you to the editor and reviewers.

Exploring crosslinguistic and genre variation in syntactic complexity: insights from a... - goo.gl/scholar/EbcSjS #ScholarAlerts

Vocabulary SIG of JALT due Jun 30 (at Kyushu Sangyo University, Fukuoka) jaltvocab.weebly.com

Towards Self-Referential Analytic Assessment: A Profile-Based Approach to L2 Writing Evaluation... - goo.gl/scholar/p2J9Bm #ScholarAlerts

Paul Nation's regular section in The Language Teacher and the expositions section in JALT Journal are excellent. Short, digestible, and making research palatable for language teachers 👏 jalt-publications.org

Copilot built into Outlook is making my life so much easier. Finding emails, summarising, checking if I am close to any deadlines, etc.

Investigating Accent Bias in AI Intelligibility of L2 English Accents - goo.gl/scholar/X7KPYs #ScholarAlerts

@nickm_jp It's probably really common isn't it 🫣

@coopersensei youtube.com/watch?v=0nCLnf… Here is some study material so we don't get left behind!

'He's so GOATed it's insane' First time I've heard GOAT used as an adjective, but apparently it's a thing

Writing the discussion section always involves using the most brain power 🧠🤯

Honoured to be invited to speak at the JALT Vocab SIG symposium in Fukuoka (26 Sept.). @MizumotoAtsushi is the invited discussant this year. There is also a call for presentations/posters out until 30 June. #responses" target="_blank" rel="nofollow noopener">docs.google.com/forms/d/1E0pLG…

It's great when conferences share videos of talks online for free, this is one from University of Cambridge sites.google.com/cam.ac.uk/wric… This talk was good: Stefano Bannò (Cambridge) Exploiting the English Vocabulary Profile for L2 word-level vocabulary assessment with LLMs.

When interests collide in a venn diagram: Mike Skinner being interviewed about brutalist architecture wepresent.wetransfer.com/stories/the-st…

@DiemerStefan Good points!

A point. I prefer general remarks as I feel reviewing is more about conveying general criticism (and praise) and leaving the revisions to the authors, but I admit that detailed feedback might make a revision easier. My concern is to not treat articles like student papers and leave room for different approaches. And to be honest it is also a matter of time allocation as a reviewer…

Is 39 comments too many to write in a peer review? Asking for a friend 😬

@DiemerStefan 😳 Maybe sometimes short, actionable comments are better than long, sweeping general ones? 😅

@coopersensei Not if you want to channel R2…

Looks interesting... Translation norms for English and Japanese translation pairs: Translation ambiguity, proficiency, and production errors link.springer.com/article/10.375…

Several useful datasets here: huggingface.co/ALTACambridge

Preparing for ICAME in Koblenz Germany from 26 May

Black marks on published papers don’t change citation rates, new study finds – Retraction Watch share.google/H7YdcRg74Gunj9…

Can LLMs flip coins in their heads? When prompted to “Flip a fair coin” 100 times, the heads to tails ratio drifts far from 50:50. LLMs can understand what the target probability should be, but generating outputs that faithfully follow a given distribution is a separate problem. This bias extends beyond coin flips. When LLMs are asked to generate multiple story ideas or brainstorm solutions, the outputs tend to cluster around a narrow range. The same probabilistic skew that distorts coin flips limits diversity in creative generation, recommendations, and other tasks where varied outputs are needed. We discovered a prompting technique named String Seed of Thought (SSoT). The method is simple: instruct the LLM to generate a random string in its own output, then manipulate that string to derive its answer. It requires only a small addition to the prompt and no external random number generator. SSoT significantly reduces output bias across a wide range of LLMs, both open and closed. With reasoning models (such as DeepSeek-R1), it reaches accuracy close to that of actual random sampling. The method generalizes from binary choices to n-way selections and arbitrary probability distributions. On the NoveltyBench diversity benchmark, SSoT outperformed other approaches across all six categories while maintaining output quality. This work will be presented at #ICLR2026! Blog: pub.sakana.ai/ssot Paper: arxiv.org/abs/2510.21150 Openreview: openreview.net/forum?id=luXtb…