7. Abstract Reasoning Composition UC San Diego and UMD propose ArcMemo, a test-time memory framework that distills reusable concepts from solution traces, stores them in natural language, and retrieves a relevant subset on future queries.
RESEARCH
-
Markovian Thinker: RL Environment for LLM State Management
By
–
6. The Markovian Thinker
— DAIR.AI (@dair_ai) 12 octobre 2025
A new RL thinking environment that keeps an LLM’s effective state constant by chunking long chains of thought and carrying over only a short textual state between chunks.https://t.co/mqXfF1XG6h6. The Markovian Thinker A new RL thinking environment that keeps an LLM’s effective state constant by chunking long chains of thought and carrying over only a short textual state between chunks.
-

Scaling Long-Horizon Reasoning Without Step Labels
By
–
5. Reasoning over Longer Horizons via RL The authors show that you can scale long-horizon reasoning without step labels or heavy scaffolding.
-

Inoculation Prompting: Safety Technique for Flawed Training Data
By
–
4. Inoculation Prompting (IP) The paper introduces a simple trick for SFT on flawed data: edit the training prompt to explicitly ask for the undesired behavior, then evaluate with a neutral or safety prompt.
-

Emergent Misalignment in Multi-Agent AI Systems
By
–
2. Emergent Misalignment In controlled multi-agent sims, models fine-tuned to maximize conversions, votes, or engagement also increased deception, disinformation, and harmful rhetoric, even when instructed to stay truthful.
-

Gene therapy restores hearing in deaf children via one-shot treatment
By
–
New @NEJM One-shot gene therapy for a type of congenital deafness enabled some children to hear for the first time, and 3 of 12 kids achieved normal hearing http://
nejm.org/doi/full/10.10
56/NEJMoa2400521
… -
Qwen 3 release questioned technical report limitations
By
–
I think the only one I might question the significance of in your list is Qwen 3. They didn't even release a base model. The technical report was quite disappointing as well.
-
ADIPEC2025 Abu Dhabi: AI Leaders Shape Energy Future
By
–
The world meets in Abu Dhabi 🇦🇪 for #ADIPEC2025 – where Energy. Intelligence. Impact. define the future.
— Giuliano Liguori (@ingliguori) 12 octobre 2025
AI leaders like Arthur Mensch (Mistral AI), Dr Guy Diedrich (Cisco) & Prof. Anima Anandkumar (Caltech) join global energy voices to shape progress.
🎥 Watch the official… pic.twitter.com/WUYu4eXbnnThe world meets in Abu Dhabi for #ADIPEC2025 – where Energy. Intelligence. Impact. define the future. AI leaders like Arthur Mensch (Mistral AI), Dr Guy Diedrich (Cisco) & Prof. Anima Anandkumar (Caltech) join global energy voices to shape progress. Watch the official
-
Inference Limited Role in AI Development Picture
By
–
Inference is a small part of the picture. Not comparable.
-

MIT Tool Visualizes and Edits Physically Impossible Objects
By
–
MIT tool visualizes and edits "physically impossible" objects https://news.mit.edu/2025/mit-meschers-tool-visualizes-edits-physically-impossible-objects-0804 @MIT
#AI #IoT #5G
