Lightnews — Scholar-powered news

Alessio Miaschi @alessiomiaschi.bsky.social · 12d

Our paper “Crossword Space: Latent Manifold Learning for Italian Crosswords and beyond” won the CLiC-it 2025 Best Student Paper Award! 🥳🥳

📚 You can check the paper at the following link: clic2025.unica.it/wp-content/u...

@ailc-nlp.bsky.social #NLProc

3

Reposted by Alessio Miaschi

Gabriele Sarti @gsarti.com · 15d

Now with sleek flyers to test your skills in Italian crossword solving! 🤗 Join our #EVALITA2026 task!

1 1

Alessio Miaschi @alessiomiaschi.bsky.social · 22d

Plenty of exciting and challenging tasks in EVALITA 2026 🚀
Check out the Call for Interest and consider participating 👉 docs.google.com/forms/d/e/1F...

#EVALITA2026 #NLPRoc

AILC-NLP @ailc-nlp.bsky.social · 23d

🚀 The Call for Interest is out!
📢 Want to participate in one of the EVALITA 2026 tasks? Check out the CFI and get ready to join the challenge 👉 www.evalita.it/campaigns/ev...

In the next few days, we’ll introduce the tasks one by one. Stay tuned!
#EVALITA2026 #NLProc #ItalianNLP

Alessio Miaschi @alessiomiaschi.bsky.social · 23d

All the details about the task are available here 👉 sites.google.com/view/crucive...
📅 Training data release: 22 September 2026

Cruciverb-IT

Overview Cruciverb-IT is the first shared task proposed at EVALITA 2026 on crossword puzzle solving. We propose two tasks: i) answering clues extracted from Italian crosswords; ii) autonomously solvin...

sites.google.com

Alessio Miaschi @alessiomiaschi.bsky.social · 23d

🚨 Exciting news from #EVALITA2026 (@ailc-nlp.bsky.social)!
I'm co-organizing Cruciverb-IT, the first shared task on crossword solving 🧩✍️ together with Ciaccio C., @gsarti.com, Dell'Orletta F. and @malvinanissim.bsky.social!
If you love cracking crosswords (or cracking models that do), join us! 🎉

1 1 2

Alessio Miaschi @alessiomiaschi.bsky.social · 27d

Had a really great time in Varna! 🇧🇬 many thanks to the organizers of the LM4DH workshop (@ranlp.bsky.social) for inviting me! 🥳 #NLProc

Alessio Miaschi @alessiomiaschi.bsky.social · Sep 2

🎉 Our work has been accepted at the EMNLP Findings 2025! 🔥

I won’t be in Suzhou, but if you are attending, make sure to check out this great work I had the pleasure of collaborating on!

#NLProc @emnlpmeeting.bsky.social #EMNLP2025

FBK - NLP Research Group @fbk-nlp.bsky.social · Sep 2

🎉 Thrilled to share that our paper “All-in-one: Understanding and Generation in Multimodal Reasoning with the MAIA Benchmark” has been accepted at the EMNLP 2025 conference! 🔥

📄 Preprint here: arxiv.org/abs/2502.16989

See you in Suzhou next November!!! 🇨🇳🚀
#EMNLP2025 #NLP #Multimodality

All-in-one: Understanding and Generation in Multimodal Reasoning with the MAIA Benchmark

We introduce MAIA (Multimodal AI Assessment), a native-Italian benchmark designed for fine-grained investigation of the reasoning abilities of visual language models on videos. MAIA differs from other...

arxiv.org

2

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 30

Today at the @aclmeeting.bsky.social main poster session, we presented: “Evaluating Lexical Proficiency in Neural Language Models” (Ciaccio C., Miaschi A. and Dell’Orletta F.)

🗒️ Paper: aclanthology.org/2025.acl-lon...

#ACL2025NLP #NLProc

1 6

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 29

Yesterday at the @aclmeeting.bsky.social Findings poster session, we presented our work “Beyond the Spelling Miracle: Investigating Substring Awareness in Character-Blind Language Models” (with Ciaccio C., Sartor M., and Dell’Orletta F.)

Paper: aclanthology.org/2025.finding...

#ACL2025NLP #NLProc

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 24

📂 Code available at the following repository: github.com/snizio/Beyon...
🧵(5/5)

GitHub - snizio/Beyond-Spelling-Miracle

Contribute to snizio/Beyond-Spelling-Miracle development by creating an account on GitHub.

github.com

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 24

✅ Larger models develop more robust substring awareness.
✅ Morphemes are recognized better than meaningless substrings.
✅ Awareness emerges early for suffixes and roots, later for non-morphemic units
✅ Productivity, word frequency and tokenization shape this ability.
🧵(4/5)

1

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 24

🧪 We design a controlled binary task asking models whether a substring appears in a word. Using MorphoLex, we evaluate models from the Pythia family across:
- substring position and length;
- morphemic vs. non-morphemic substrings;
- pre-training checkpoints.
🧵(3/5)

1

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 24

LMs operate on subword tokens and lack explicit access to characters. Despite so, they show a limited ability to recognize spelling-level patterns (i.e. Spelling Miracle). In this work, we take a look at when, where, and how such character-level awareness emerges.
🧵(2/5)

1 1 1

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 24

🚀 On Monday 28th, we will present our #ACL2025NLP (@aclmeeting.bsky.social) Findings Paper "Beyond the Spelling Miracle: Investigating Substring Awareness in Character-Blind Language Models" (with Ciaccio C., Sartor M. and Dell'Orletta F.)

🔗 aclanthology.org/2025.finding...
🧵(1/5)

#NLProc

1 1

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 23

🤗 Models available on Huggingface: huggingface.co/collections/...
📂 Code & Dataset: github.com/snizio/Lexic...
🧵(5/5)

Evaluating Lexical Proficiency in Neural Language Models - a snizio Collection

Public collection for our paper: "Evaluating Lexical Proficiency in Neural Language Models", C. Ciaccio, A. Miaschi, F. Dell'Orletta (ACL 2025)

huggingface.co

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 23

🧠 Our findings show that Transformer-based models can handle lexical composition and meaning inference to some extent—effectively producing and interpreting plausible lexical innovations, though with a notable drop in performance vs. standard lexical items.
🧵(4/5)

1

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 23

Key contributions:
✅ A new framework to assess lexical abilities across tasks & word types
✅ A lexical resource for Italian with definitions & examples
✅ Analysis of model size, multilinguality & linguistic features
✅ Human eval via the Optimal Innovation Hypothesis
🧵(3/5)

1

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 23

In this study, we propose a novel, unified framework to evaluate lexical proficiency in Transformer-based LMs, testing their ability to generate, define, and use words across three lexical categories: commonly lexicalized words, recent neologisms and nonce words.
🧵(2/5)

1

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 23

📣 Next week I’ll be at @aclmeeting.bsky.social with three papers: one at the main conference and two in the Findings!

At the main conference, I’ll present:
“Evaluating Lexical Proficiency in Neural Language Models” (with Ciaccio C. and Dell’Orletta F.)
🔗 aclanthology.org/2025.acl-lon...
🧵(1/5)

1

Alessio Miaschi @alessiomiaschi.bsky.social · Jul 21

Just one week left to submit your task proposal for #EVALITA2026!
Deadline: Monday, 28th July 🕐🕐
Don't miss the chance to be part of the evaluation campaign! 🥳 #NLProc @ailc-nlp.bsky.social

AILC-NLP @ailc-nlp.bsky.social · Jun 19

The second #CFP of EVALITA 2026 is out and published in the workshop website 🌍 www.evalita.it/campaigns/ev... 📝 July 28th 2025 (extended!): submission of task proposals 🏆 August 7th 2025 (extended!): notification of task proposal acceptance 🇮🇹 #EVALITA2026 #NLProc

EVALITA 2026: Second call for tasks

EVALITA 2026: Second call for tasks NEW DEADLINES AND TIMELINE EVALITA 2026 is an initiative of AILC (Associazione Italiana di Linguistica Computazionale). As in the previous editions, EVALITA 2026 ...

www.evalita.it

1

Alessio Miaschi @alessiomiaschi.bsky.social · Jun 19

The second #CFP of #EVALITA2026 is out! 🔥 Deadline for submitting task proposals is July 28th 2025!
#NLProc

AILC-NLP @ailc-nlp.bsky.social · Jun 19

The second #CFP of EVALITA 2026 is out and published in the workshop website 🌍 www.evalita.it/campaigns/ev... 📝 July 28th 2025 (extended!): submission of task proposals 🏆 August 7th 2025 (extended!): notification of task proposal acceptance 🇮🇹 #EVALITA2026 #NLProc

EVALITA 2026: Second call for tasks

EVALITA 2026: Second call for tasks NEW DEADLINES AND TIMELINE EVALITA 2026 is an initiative of AILC (Associazione Italiana di Linguistica Computazionale). As in the previous editions, EVALITA 2026 ...

www.evalita.it

Alessio Miaschi @alessiomiaschi.bsky.social · Jun 3

🚀 Our latest paper has been accepted to Findings of #ACL2025! Check it out here: arxiv.org/pdf/2505.24523
@aclmeeting.bsky.social #NLProc

Michele Papucci @mpapucci.bsky.social · Jun 3

🧵1/ Machine-Generated Text (MGT) detection is failing

Our paper, accepted at Findings of ACL 2025, shows that LLMs can fool generated-text detectors.
arxiv.org/abs/2505.24523

Andrea Pedrotti, Cristiano Ciaccio, @alessiomiaschi.bsky.social @gpucce.bsky.social, Felice Dell'Orletta, Adrea Esuli

5

Alessio Miaschi @alessiomiaschi.bsky.social · May 16

🔜 More info coming soon!

Alessio Miaschi @alessiomiaschi.bsky.social · May 16

3) Findings: Stress-testing Machine Generated Text Detection: Shifting Language Models Writing Style to Fool Detectors (with Pedrotti A., Papucci M., Ciaccio C., Puccetti G., Dell'Orletta F. and Esuli A.)

1 3

Alessio Miaschi @alessiomiaschi.bsky.social · May 16

2) Findings: Beyond the Spelling Miracle: Investigating Substring Awareness in Character-Blind Language Models (with Ciaccio C., Sartor M. and Dell'Orletta F.)

1 2