In an environment of perpetual digital noise, geopolitical friction, and algorithmic manipulation, many users have lost faith in the integrity of the information they see. The information landscape is saturated, blurring the lines between ordinary discourse and strategic misinformation by companies or states. If high-stakes decisions (from investment strategies to international security choices) are based on data that can be fabricated or perpetually disputed, the global system will face a crisis of integrity and legitimacy. One way to […]

Ver mais

(The post was originated from my talk for WiMLDS x Fintech meetup hosted by Affirm.) I believe many of you have watched or heard of the games between AlphaGo and professional Go player Lee Sedol in 2016. Lee has the highest rank of nine dan and many world championships. No doubt, he is one of the best Go players in the world, but he lost by 1-4 in this series versus AlphaGo. Before this, Go was considered to […]

Ver mais

arXiv:2503.09887v3 Announce Type: replace-cross Abstract: We develop a novel stability theory for Sinkhorn semigroups based on Lyapunov techniques and quantitative contraction coefficients, and establish exponential convergence of Sinkhorn iterations on weighted Banach spaces. This operator-theoretic framework yields explicit exponential decay rates of Sinkhorn iterates toward Schr”odinger bridges with respect to a broad class of $phi$-divergences and Kantorovich-type distances, including relative entropy, squared Hellinger integrals, $alpha$-divergences, weighted total variation norms, and Wasserstein distances. To the best of our knowledge, […]

Ver mais

arXiv:2602.12300v1 Announce Type: new Abstract: We propose a new method, that we coined the “morphism-trick”, to integrate custom implementations of vector-Jacobian products in automatic differentiation softwares, applicable to a wide range of semiring-based computations. Our approach leads to efficient and semiring-agnostic implementations of the backward pass of dynamic programming algorithms. For the particular case of finite-state methods, we introduce an algorithm that computes and differentiates the $oplus$-sum of all paths’ weight of a finite-state automaton. Results show that, […]

Ver mais

arXiv:2601.16400v1 Announce Type: new Abstract: Real-world visual question answering (VQA) is often context-dependent: an image-question pair may be under-specified, such that the correct answer depends on external information that is not observable in the image. In such cases, directly answering can lead to confident but incorrect predictions. We propose CoA(Clarify-or-Answer), an ask-or-answer agent that separately models the decision to ask or answer, and what to ask if needed. CoA first determines whether clarification is necessary; if so, it […]

Ver mais

For years, the future of the internet was sold to us as a speed problem. Faster downloads. Lower latency. Better streaming. More connected devices. Smarter cities. Smarter homes. Smarter everything. That promise gave 5G a simple story to tell. The internet people already knew would become quicker, smoother, and more powerful. In many ways, that story made sense. Speed matters. Latency matters. Infrastructure matters. The modern digital world runs on responsiveness, and every new layer of connectivity changes […]

Ver mais

arXiv:2603.22563v1 Announce Type: new Abstract: Preference-based fine-tuning has become an important component in training large language models, and the data used at this stage may contain sensitive user information. A central question is how to design a differentially private pipeline that is well suited to the distinct structure of reinforcement learning from human feedback. We propose a privacy-preserving framework that imposes differential privacy only on reward learning and derives the final policy from the resulting private reward model. […]

Ver mais

arXiv:2602.05230v1 Announce Type: cross Abstract: Linear attention methods offer Transformers $O(N)$ complexity but typically underperform standard softmax attention. We identify two fundamental limitations affecting these approaches: the restriction to convex combinations that only permits additive information blending, and uniform accumulated weight bias that dilutes attention in long contexts. We propose Zero-Sum Linear Attention (ZeroS), which addresses these limitations by removing the constant zero-order term $1/t$ and reweighting the remaining zero-sum softmax residuals. This modification creates mathematically stable weights, […]

Ver mais