Hello everyone, I wanted to share a project I am working on for a hackathon. It’s a reinforcement learning environment where an AI agent acts as a power grid operator. I’ve tried to keep physics and maths as real as possible. Github repo link : https://github.com/krishnagoyal099/Opengrid_env Live link : https://huggingface.co/spaces/K446/Opengrid I would really like to get your feedback on the physics modeling and reward structure, and also if anyone manages to solve the “hard” task! I am willing […]

Ver mais

Reinforcement Learning with Verifiable Rewards (RLVR) has become a prominent method for post-training Large Language Models (LLMs). However, verifiers are rarely error-free; even deterministic checks can be inaccurate, and the growing dependence on model-based judges exacerbates the issue. The extent to which RLVR is robust to such noise and the verifier accuracy required for effective training remain unresolved questions. We investigate these questions in the domains of code generation and scientific reasoning by introducing noise into RL training. […]

Ver mais

WebRTC is designed to degrade and drop my prompt during poor network conditions. wtf my dude WebRTC aggressively drops audio packets to keep latency low. If you’ve ever heard distorted audio on a conference call, that’s WebRTC baybee. The idea is that conference calls depend on rapid back-and-forth, so pausing to wait for audio is unacceptable. …but as a user, I would much rather wait an extra 200ms for my slow/expensive prompt to be accurate. After all, I’m […]

Ver mais

arXiv:2604.09734v1 Announce Type: new Abstract: We study how far structured architectural bias can compensate for the absence of end-to-end gradient-based representation learning in visual recognition. Building on the VisNet tradition, we introduce a modular hierarchical framework combining: (i) fixed multi-frequency Gabor decomposition into F=7 parallel streams; (ii) within-stream competitive learning with Hebbian and Oja updates and anti-Hebbian decorrelation; (iii) an associative memory module inspired by modern Hopfield retrieval; and (iv) iterative top-down modulation using local prediction and reconstruction […]

Ver mais

arXiv:2603.06618v1 Announce Type: new Abstract: Multiplex Biological Networks (MBNs), which represent multiple interaction types between entities, are crucial for understanding complex biological systems. Yet, existing methods often inadequately model multiplexity, struggle to integrate structural and sequence information, and face difficulties in zero-shot prediction for unseen entities with no prior neighbourhood information. To address these limitations, we propose a novel framework for zero-shot interaction prediction in MBNs by leveraging context-aware representation learning and knowledge distillation. Our approach leverages domain-specific […]

Ver mais

arXiv:2603.10047v1 Announce Type: new Abstract: Hallucinations in large language models (LLMs) are outputs that are syntactically coherent but factually incorrect or contextually inconsistent. They are persistent obstacles in high-stakes industrial settings such as engineering design, enterprise resource planning, and IoT telemetry platforms. We present and compare five prompt engineering strategies intended to reduce the variance of model outputs and move toward repeatable, grounded results without modifying model weights or creating complex validation models. These methods include: (M1) Iterative […]

Ver mais

arXiv:2504.14094v3 Announce Type: replace-cross Abstract: Concept-based Models aim to improve interpretability by predicting high-level intermediate concepts, representing a promising approach for deployment in high-risk scenarios. However, they are known to suffer from information leakage, whereby models exploit unintended information encoded within the learned concepts. We introduce an information-theoretic framework to rigorously characterise and quantify leakage, and define two complementary measures: the concepts-task leakage (CTL) and interconcept leakage (ICL) scores. We show that these measures are strongly predictive of […]

Ver mais

RL is ridiculously cool when the state space is huge or the dynamics unknown. But for a large class of problems where you do have a model, the state is relatively small (<8 dims or so?), and you want an exact policy across the entire support, backward induction on a grid is still a fast and viable option. I couldn’t find a good implementation of this using a GPU backend, so I built bellgrid. It’s a PyTorch-based DP […]

Ver mais

My perfectly stable homelab server gradually descended into chaos — random crashes, segfaults, system hangs — and how months of troubleshooting, community support, and hardware swaps eventually led to one conclusion: the CPU itself was defective from the factory. If you’re experiencing unexplained instability on an Intel 13th or 14th Gen system, this story might save you months of frustration. Quick Symptom Checklist If you’re seeing multiple of these symptoms, you may be affected: Random segfaults in unrelated […]

Ver mais