digitado

In this work, we study whether enforcing strict compositional structure in sequence embeddings yields meaningful geometric organization when applied to protein-protein interaction networks. Using Event2Vec, an additive sequence embedding model, we train 64-dimensional representations on random walks from the human STRING interactome, and compare against a DeepWalk baseline based on Word2Vec, trained on the same walks. We find that compositional structure substantially improves pathway coherence (30.2$times$ vs 2.9$times$ above random), functional analogy accuracy (mean similarity 0.966 vs 0.650), […]

Ver mais

Learning to defer (L2D) enables human-AI cooperation by deciding when an AI system should act autonomously or defer to a human expert. Existing L2D methods, however, assume static human performance, contradicting well-established findings on fatigue-induced degradation. We propose Fatigue-Aware Learning to Defer via Constrained Optimisation (FALCON), which explicitly models workload-varying human performance using psychologically grounded fatigue curves. FALCON formulates L2D as a Constrained Markov Decision Process (CMDP) whose state includes both task features and cumulative human workload, and […]

Ver mais

Medical imaging techniques, especially Magnetic Resonance Imaging (MRI), are accepted as the gold standard in the diagnosis and treatment planning of neurological diseases. However, the manual analysis of MRI images is a time-consuming process for radiologists and is prone to human error due to fatigue. In this study, two different Deep Learning approaches were developed and analyzed comparatively for the automatic detection and classification of brain tumors (Glioma, Meningioma, Pituitary, and No Tumor). In the first approach, a […]

Ver mais

Reinforcement Learning with Verifiable Rewards (RLVR) has become a central post-training paradigm for improving the reasoning capabilities of large language models. Yet existing methods share a common blind spot: they optimize policies based on instantaneous group-level or batch-level statistics without ever verifying whether the resulting update actually improved the model. This open-loop design — updating in isolation at each step, guided only by within-group (batch) reward signals — means optimization can drift or collapse with no mechanism to […]

Ver mais

Following up on some thoughts around RLHF and LLM training. Most current RLHF pipelines can be framed as optimizing a policy πₜ (the LLM) against a learned reward model r_φ that approximates human preference distributions over outputs. In practice, this is often implemented with PPO-style updates under KL constraints relative to a reference policy. This setup works well for alignment and helpfulness, but it has a few structural properties that seem limiting: 1. Static reward modeling The reward […]

Ver mais

Hi, I have been studying RL for the last 3 months. I wanted to create a project on a recommendation system. I have been a little lost on this path and wanted to ask for suggestions on the following: 1) Any basic research papers that I should read that can describe the process and problems faced? 2) Any beginning structure that you would recommend? 3) Any thoughts on problems like cold starts? 4) Anything else that you would […]

Ver mais