As datasets continue to expand in size and complexity, these models have become increasingly sophisticated, with deeper architectures and greater expressive power. Despite these advances, DNNs trained on imbalanced class distributions often exhibit a tendency to favor majority classes, leading to degraded performance on underrepresented classes [18, 39, 27, 17]. Because many real-world datasets follow long-tailed distributions in which minority classes can contain critical and informative patterns, developing methods that enable DNNs to learn effectively from imbalanced data is essential to prevent the loss of valuable information from these rare classes [26, 34, 16]. Moreover, data encountered in real-world applications are frequently multi-modal, meaning that observations originate from heterogeneous sources [6, 29, 7, 35]. To make effective use of such heterogeneous inputs, a wide range of multi-modal learning approaches have been proposed that exploit complementary information across modalities to enhance predictive performance [10, 5]. Common strategies integrate multiple modalities into a unified representation, using techniques that span from straightforward feature-level concatenation [19, 11, 12] to more sophisticated neural architectures that learn joint representations in an end-to-end manner [20, 32]. Although prior research has extensively studied class imbalance and multi-modal data separately, relatively little attentionhas beengiven to settings where bothchallenges arise si2 multaneously. Developing methods that can effectively handle long-tailed class distributions in conjunction with multi-modal inputs is therefore essential in many real-world applications. In the medical domain, for instance, datasets often contain far more samples from healthy individuals than from patients with specific conditions, while also encompassing diverse datatypes such asimagingdata(e.g., X-rays)alongsideauxiliary informationincluding demographics and clinical histories.

Large language models are increasingly deployed with test-time strategies: sample $N$ responses, score them with a reward model or verifier, and return the best. This deployment rule exposes a mismatch in post-training: standard objectives optimize the mean reward of a single response, whereas best-of-$N$ performance is governed by the upper tail of the reward distribution. Recent test-time-aware objectives partly address this mismatch, but typically assume that training can use the same per-prompt rollout budget as deployment, which is impractical when post-training must cover many prompts while deployment can allocate much larger per-prompt test-time compute. We study this budget-mismatch regime, where only $m\ll N$ per-prompt rollouts are available during training but the target objective is best-of-$N$ deployment. Under structural assumptions on the reward tails, we show that the policy gradient of the best-of-$N$ objective can be approximated from a much smaller rollout group by extrapolating upper-tail statistics. This yields a family of Tail-Extrapolated estimators for best-of-$N$-oriented post-training: a simple direct estimator, Tail-Extrapolated Advantage (TEA), and a fixed-order debiased Prefix-TEA estimator based on moment cancellation. Experiments on instruction-following tasks show that TEA and Prefix-TEA improve best-of-$N$ performance across different language models, reward models and datasets under various training and test-time budget settings.

Ilya Sutskever Stands by His Role in Sam Altman's OpenAI Ouster: 'I Didn't Want It to Be Destroyed' The former OpenAI chief scientist may be estranged from the company, but he still came to its defense as he testified on Monday. Elon Musk's trial against OpenAI and Microsoft entered its final stretch on Monday, with testimony from Microsoft CEO Satya Nadella, former OpenAI chief scientist Ilya Sutskever, and current OpenAI chairman Bret Taylor. Sutskever drew the spotlight, revealing an ownership stake in OpenAI's $850-billion for-profit arm that is currently worth about $7 billion. That makes him one of the largest known individual shareholders of OpenAI. Earlier in the trial, OpenAI president Greg Brockman acknowledged for the first time that he has around $30 billion worth of OpenAI shares .

Daron Acemoglu is more cautious than most about predictions of a jobs apocalypse. A few months before he was awarded the Nobel Prize in economics in 2024, Daron Acemoglu published a paper that earned him few fans in Silicon Valley. Contrary to what Big Tech CEOs had been promising--an overhaul of all white-collar work--Acemoglu estimated that AI would give only a small boost to US productivity and would not obviate the need for human work. It's okay at automating certain tasks, he wrote, but some jobs will be perfectly fine. Two years later, Acemoglu's measured take has not caught on. Chatter about an AI jobs apocalypse pops up everywhere from Senator Bernie Sanders's rallies to conversations I overhear in line at the grocery store.

Plus: Meta's embrace of AI is making employees miserable. Here's what you need to know about the cruise ship hantavirus outbreak Last week, eight passengers aboard a Dutch-flagged cruise ship contracted a type of hantavirus transmitted by rats. But health experts stress that this situation is nothing like the coronavirus outbreak in 2020. The Andes virus is known to spread between people, and there are no specific antiviral treatments or vaccines. Yet transmission appears to require a specific form of contact that the cruise ship fostered. Here's what you need to know about the outbreak--and why experts believe it can be contained .

There's a deep, forbidding moat that surrounds Nvidia--and it has nothing to do with hardware. Forgive me for starting with a cliché, a piece of finance jargon that has recently slipped into the tech lexicon, but I'm afraid I must talk about "moats." Popularized decades ago by Warren Buffett to refer to a company's competitive advantage, the word found its way into Silicon Valley pitch decks when a memo purportedly leaked from Google, titled "We Have No Moat, and Neither Does OpenAI," fretted that open-source AI would pillage Big Tech's castle. A few years on, the castle walls remain safe. Apart from a brief bout of panic when DeepSeek first appeared, open-source AI models have not vastly outperformed proprietary models.

We study the spectral properties of sample covariance matrices constructed from pooled sequence representations, where token embeddings are drawn from a fixed two-class Gaussian mixture table and pooled via (fixed) attention weights. Working in the high-dimensional regime $d,V,N\to\infty$ with $d/V\toδ$ and $d/N\toγ$, we derive exact characterizations of the limiting eigenvalue distribution, outlier eigenvalues, and eigenvector alignment with the hidden signal. The bulk spectrum follows a non-Marchenko--Pastur law given by the free multiplicative convolution $κ(MP_δ\boxtimes MP_γ)$, reflecting the finite vocabulary structure. Signal recovery undergoes two successive BBP-type phase transitions characterized by the scalars: $δ,γ,α=w^{\top} R w$ and $κ=\|w\|^2$, where $w$ denotes the attention pooling weights and $R$ the positional correlation matrix. An aftermath of our analysis demonstrates that the optimal attention weights maximizing the signal-to-noise ratio $α/κ$ are given by the (normalized) top eigenvector of $R$, and we show (as a particular case of our analysis) that parameter-free causal self-attention with $τ/d$ score scaling yields deterministic harmonic weights that improve signal recovery over mean pooling whenever early tokens carry more signal. Extensive simulations confirm sharp agreement between theory and finite-dimensional experiments.

Reinforcement Learning from Human Feedback (RLHF) has become a cornerstone technique for post-training large language models. While most existing approaches rely on the reverse KL-regularization, recent empirical studies have begun exploring alternative divergences (e.g., forward KL, chi-squared) as regularizers in RLHF. However, a unified theoretical understanding of general $f$-divergence regularization remains under-explored. To fill this gap, this work develops a comprehensive theoretical framework for online RLHF with a general $f$-divergence regularized objective. Rather than treating each possible divergence function individually, we adopt a holistic perspective across the entire function class and propose two algorithms based on distinct sampling principles. The first extends the classical optimism principle with a carefully designed exploration bonus, while the second introduces a new method that exploits the sensitivity of the optimal policy to reward perturbations under $f$-divergence regularization. Theoretical analysis shows that $O(\log T)$ regret and $O(1/T)$ sub-optimality gap are achievable, establishing provable efficiency of both algorithms and, to the best of our knowledge, the first performance bounds for online RLHF under general $f$-divergence regularization.

Aligning large language models (LLMs) to human preferences typically relies on aggregating pooled feedback into a single reward model. However, this standard approach assumes that all labelers share the same underlying preferences, ignoring the fact that real-world labelers are highly heterogeneous and usually anonymous. Consequently, relying solely on binary choice data fundamentally distorts the learned policy, making the true population-average preference unidentifiable. To overcome this critical limitation, we demonstrate that augmenting preference datasets with a simple, secondary signal -- the user's response time -- can restore the identifiability of the population's average preference. By modeling each decision as a Drift-Diffusion Model (DDM), we introduce a novel, consistent estimator of heterogeneous preferences that successfully corrects the distortions of standard choice-only labels. We prove that our estimator asymptotically converges to the true average preference even in extreme cases where each anonymous labeler contributes only a single choice. Empirically, across both synthetic and real-world datasets, our method consistently outperforms standard baselines that otherwise fail and plateau at a bias floor. Because response times are essentially free to record and require zero user tracking or identification, our results bring promises and open up new opportunities for future data-collection pipelines to improve the social benefit without requiring user-level identifiers or repeated elicitations.

AI agents are increasingly deployed in multi-task settings, where the task to perform is specified at test time, and the agent must generalize to unseen tasks. A major concern in such settings is safety: often, an agent must not only execute unseen tasks, but do so while avoiding risks and handling ones that materialize. Empirical evidence suggests that even when the ability to execute generalizes to unseen tasks, the ability to do so safely frequently does not. This paper provides theory and experiments indicating that failures of agentic safety to generalize across tasks are not merely due to limitations of training methods, but reflect an inherent property of safety itself: the relationship between a task and its safe execution is more complex than the relationship between a task and its execution alone. Theoretically, we analyze linear-quadratic control with $H_{\infty}$-robustness, and prove that the mapping from task specification to an optimal controller has higher Lipschitz constant with safety requirements than without, yielding a Lipschitz bound of independent interest. Empirically, we demonstrate our conclusions in simulated quadcopter navigation with a neural network agent and in CRM with an LLM agent. Our findings suggest that current efforts to enhance agentic safety may be insufficient, and point to a need for fundamentally different approaches.