Sparse autoencoders (SAEs) decompose language model activations into interpretable features, but existing methods reveal only which features activate, not which change model outputs when amplified. We introduce Control Reinforcement Learning (CRL), which trains a policy to select SAE features for steering at each token, producing interpretable intervention logs: the learned policy identifies features that change model outputs when amplified. Adaptive Feature Masking encourages diverse feature discovery while preserving singlefeature interpretability. The framework yields new analysis capabilities: branch point […]

Ver mais

arXiv:2602.06584v1 Announce Type: cross Abstract: Standard chain-of-thought reasoning generates a solution in a single forward pass, committing irrevocably to each token and lacking a mechanism to recover from early errors. We introduce Inference-Time Rethinking, a generative framework that enables iterative self-correction by decoupling declarative latent thought vectors from procedural generation. We factorize reasoning into a continuous latent thought vector (what to reason about) and a decoder that verbalizes the trace conditioned on this vector (how to reason). Beyond […]

Ver mais

We study the problem of state representation learning for control from partial and potentially high-dimensional observations. We approach this problem via cost-driven state representation learning, in which we learn a dynamical model in a latent state space by predicting cumulative costs. In particular, we establish finite-sample guarantees on finding a near-optimal representation function and a near-optimal controller using the learned latent model for infinite-horizon time-invariant Linear Quadratic Gaussian (LQG) control. We study two approaches to cost-driven representation learning, […]

Ver mais

arXiv:2602.23411v1 Announce Type: new Abstract: The structural phase transitions and computational complexity of random 3-SAT instances are traditionally described using thermodynamic analogies from statistical physics, such as Replica Symmetry Breaking and energy landscapes. While providing profound macroscopic insights, these theories lack a discrete microscopic structure. In this paper, we propose a complementary, strictly discrete geometric model that maps these phenomena directly to the combinatorial topology of an $N$-dimensional Boolean hypercube. By defining the problem space purely through valid […]

Ver mais

In hierarchical multi-label classification, a persistent challenge is enabling model predictions to reach deeper levels of the hierarchy for more detailed or fine-grained classifications. This difficulty partly arises from the natural rarity of certain classes (or hierarchical nodes) and the hierarchical constraint that ensures child nodes are almost always less frequent than their parents. To address this, we propose a weighted loss objective for neural networks that combines node-wise imbalance weighting with focal weighting components, the latter leveraging […]

Ver mais

Email marketing only works if messages reach the inbox. Deliverability testing identifies spam triggers, broken links, and authentication issues. Tools like MailGenius help marketers optimize campaigns, improve engagement, protect sender reputation, and maximize ROI by ensuring emails are seen, not lost to spam or promotions tabs.

Ver mais

We study uplift estimation for combinatorial treatments. Uplift measures the pure incremental causal effect of an intervention (e.g., sending a coupon or a marketing message) on user behavior, modeled as a conditional individual treatment effect. Many real-world interventions are combinatorial: a treatment is a policy that specifies context-dependent action distributions rather than a single atomic label. Although recent work considers structured treatments, most methods rely on categorical or opaque encodings, limiting robustness and generalization to rare or newly […]

Ver mais

Most founders have a version of this story. Personally, I’ve watched this happen (and made this mistake myself): the tool looks great in a demo, then headcount grows and reality hits. You pick a modern, impressive-looking tool. The demo is smooth. The UI feels fast. Everyone’s excited. For a while, it works. Then the company grows. You hire more people. Someone leaves. Finance asks for historical data. An auditor wants proof of approvals. A manager needs limited access. […]

Ver mais