#AAAI2026 #ProcessSupervision #Reasoning #RewardModelling #ReferenceGuidedEvaluation #NLP #NLProc #LLMs

Here’s Pedro at yet another international conference! 🙌✨
GAIPS member Pedro P. Santos presented “Centralized training with hybrid execution in multi-agent reinforcement learning via predictive observation imputation” at #AAAI2026, Singapore 🇸🇬
📄 Check out his paper: doi.org/10.1016/j.ar...

Thanks to everyone who contributed to a productive @aaai.org conference in Singapore. Our team enjoyed the conversations with researchers and practitioners advancing AI. See you next year! #AAAI2026

This framework, Compare&Generate,
at #AAAI2026 workshop tackles a major hurdle in synthetic data: quality control. Instead of just "thinking step-by-step," the model learns why one output is better than another to iteratively improve. @aaai.org

Attending the #FAST workshop at @aaai.org #AAAI2026. Looking forward to insightful conversations about how to build #IntelligentAgents in the era of large foundation models.

If you are interested in reinforcement learning or reinforcement learning training of language models, this might spark your interest!

✨See my new paper on scaling, capacity and complexity of reinforcement learning published at #AAAI2026 ! @aaai.org

#AAAI #AAAI26 #ReinforcementLearning #DeepRL

This #AAAI2026 workshop talk on a new distributional treatment for time series anomaly detection is a paradigm shift—using Isolation Distributional Kernels (IDK). Fascinating to see how IDK^2 maps points through Hilbert spaces to detect group anomalies. @aaai.org

at #AAAI2026 on "RECORd: A Multi-Agent LLM Framework for Reverse Engineering Codebase to Causal Relational Diagram." 👩‍💻🔍
By using reinforcement fine-tuning (RFT) and multi-agent systems, this work transforms complex code into interpretable causal graphs. @aaai.org

"Lexicographic Bandits" at #AAAI2026! 🎰bridging the gap between Regret Minimization and Best Arm Identification. In complex decision-making systems where objectives have a strict priority, finding the optimal balance is a tough challenge. @aaai.org

“Flash Embeddings for Online Learning of Categorical Features” at #AAAI2026
Fascinating approach to learning high-cardinality and recurring categorical data over time—with fixed memory constraints. @aaai.org

Just attended a fascinating talk on "Bandit Learning in Housing Markets" at #AAAI2026 🏠📊 @aaai.org

Combining matching theory with multi-armed bandits to learn core allocations—both centralized and decentralized settings with provable regret bounds.

🚨 Spotlight Talk @ AAAI 2026

Our work on biologically interpretable pattern separation using unsupervised Hebbian learning was selected as a spotlight talk at the AAAI workshop on Neuro for AI & AI for Neuro 🎉

🗓 Jan 27 · 📍 Singapore EXPO

#AAAI2026 #NeuroAI #CompNeuro

Shift toward using LLMs for both predictions and natural-language explanations is a game-changer for transparency. Seeing impressive results from "TWICE-Rec" on generating high-quality rationales across diverse domains like fashion and scientific research. #AAAI2026 @aaai.org

Why do LLMs give unreliable recommendations? Often because we lose "evidence strength" during normalization. GUIDER leverages LLM logits to quantify uncertainty and uses a dynamic re-ranking strategy to boost transparency and trust. #AAAI2026 @aaai.org

Why State-Space Models for event cameras? MA-Mamba proves it! By integrating a Spatio-Temporal Association module, they’ve solved the "noisy & inconsistent" channel update issue in standard SSMs. Great results on DSEC and MVSEC #AAAI2026 @aaai.org

Deep dive into Hyperspectral Image SR today at #AAAI2026. The GEWDiff model uses an edge-aware EDM noise scheduler and a multi-level loss function to ensure structural invariance and stable convergence. Really clean results compared to other SOTA models. @aaai.org

Fascinated by the "Think-Free" ranking approach (TFRank) presented at #AAAI2026! 🚀 It internalizes complex reasoning into small LLMs (<10B), achieving high-accuracy document ranking without the latency of explicit CoT. Practical, efficient, and very impressive. @aaai.org

watched Ed Feigenbaum talk at #AAAI2026 and apparently the way he got into AI research was kinda like mine --- got bored as an elec. eng. freshman, took a class on "Ideas & Social Change", which led to studying with Herbert Simon on Math. Models in the Social Sciences, and hence, AI!

Real-time Novel View Synthesis with just 4 cameras? 🤯

Checked out the PHOTONS demo by China Telecom at #AAAI2026. It achieves 2K resolution at 25 FPS using a sparse view setup (only 4 RGB cameras).qq.com @aaai.org

Ready to level up your career? 🚀 Heading to the Job Fair #AAAI2026 today! If you're looking for new opportunities or a complete career pivot, this is the place to be. Time to network and make things happen! 💼✨ @aaai.org

#AAAI2026 5K Fun Run at 6am 🙋‍♀️ Can't believe I actually saw the Merlion without its water fountain! @aaai.org

I'll be presenting this work on January 25th (Hall 2, poster 41) at #AAAI2026 in Singapore!

Please stop by and reach out if you'd like to chat 😁

🎉 Excited to share our new paper which was accepted to #AAAI2026!

As LLMs become increasingly used as sources of factual knowledge, we ask:
Do they perform equitably across users of different backgrounds?

🧵⬇️
1/6

High-quality Tabular Data Synthesis with Limited Samples! 🚀The team from The University of Melbourne introduced CtrTab at #AAAI2026 It enhances diffusion models with a novel control mechanism to handle high-dimensional data constraints.📊 @aaai.org

Why treat Time Series as just numbers? 📉🤔

Fascinating talk at #AAAI2026 on using VLMs for Anomaly Detection：1️⃣ TS models lack 'world knowledge' (context). 2️⃣ LLMs lack 'visual understanding' of shapes.
@aaai.org
Solution: Plot the time series as images!

How do we reconcile Autoregressive vs. One-shot generative paradigms in Time Series? 🤔The TimeCAP presentation at #AAAI2026 offers a solution: during fine-tuning, we need to balance these complementary strengths while properly modeling multivariate dependencies. @aaai.org

2nd Invited Talk from Dr. Yolanda Gil at #AAAI2026 day2 🌟 can not agree more: AI should make us better people! @aaai.org

2nd Invited Talk from Dr. Yolanda Gil at #AAAI2026 day2 🌟 can not agree more: AI should make us better people! @aaai.org

As a #Zootopia fan, this poster at #AAAI2026 is really appealing🤩！🐰🦊 @aaai.org Very creative design and make me thinking about how to attract other's when making posters

Found a gem for the GP & Causal Inference crowd at #AAAI2026! 💎 'CaDyT'—a new framework for Causal Structure Learning in Dynamical Systems. It cleverly uses GP to model continuous-time dynamics from irregularly sampled data, using an MDL score to find causal graph @aaai.org