GFlowNets for AI-driven scientific discovery
PulseAugur coverage of GFlowNets for AI-driven scientific discovery — every cluster mentioning GFlowNets for AI-driven scientific discovery across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
GFlowNets used to generate novel LLM attacks in English and Turkish
Researchers have developed a novel method using GFlowNets to automatically generate adversarial attacks against Large Language Models (LLMs). This approach trains an attacker model to identify vulnerabilities in a victi…
-
GFlowNet interpretability study questions learned chemistry in drug discovery models
A new study published on arXiv investigates the interpretability of GFlowNets, a type of AI model used for drug discovery. Researchers developed a framework to analyze SynFlowNet, a GFlowNet trained on drug-likeness, an…
-
GFlowNet training methods explored via policy gradients and information geometry · 2 sources tracked
Two new research papers explore advanced training methods for Generative Flow Networks (GFlowNets). The first paper introduces a policy-gradient-based framework that bridges GFlowNet's flow balance with reinforcement le…
-
New AlphaG-OPD framework enhances symbolic alpha factor discovery
Researchers have developed AlphaG-OPD, a new framework for symbolic alpha factor discovery that enhances the guidance provided by generative flow networks. This method addresses the limitation of existing models by offe…
-
New regression losses enhance Generative Flow Network training
Researchers have developed new regression loss functions for Generative Flow Networks (GFlowNets) to improve their training process. By theoretically linking regression losses to specific divergence measures, the team d…
-
New TILDE method enables concept unlearning in text-to-image models
Researchers have developed TILDE (TILt-based Distributional Erasure), a new method for concept unlearning in text-to-image diffusion models. This technique addresses the challenge of removing specific concepts, such as …
-
New GFlowNet framework enhances active learning for molecular discovery
Researchers have developed a new active learning framework called BALD-GFlowNet, which utilizes Generative Flow Networks (GFlowNets) to improve the scalability of active learning, particularly for large datasets in area…
-
New MCMC method uses neural nets to adaptively stop sampling
Researchers have developed a new framework that uses neural classifiers to adaptively determine when to stop sampling in Markov chain Monte Carlo (MCMC) methods. This approach, framed within Generative Flow Networks (GF…
-
Proximal Policy Optimization Enhances GFlowNet Training
Researchers have introduced Proximal Policy Optimization (PPO) as a novel method for training Generative Flow Networks (GFlowNets). This approach leverages connections between GFlowNets and entropy-regularized reinforce…
-
New GFlowNet method generates highly synthesizable molecules
Researchers have developed a new method called S3-GFN for generating molecules that are both synthesizable and possess desirable properties. This approach uses a sequence-based Generative Flow Network (GFlowNet) with so…
-
New GFlowNet Framework Composes Pre-trained Models for Multi-Objective Generation
Researchers have developed a new framework for Generative Flow Networks (GFlowNets) that allows for the composition of pre-trained models at inference time. This approach enables rapid adaptation to new multi-objective …
-
New GFlowNet training method improves LLM prefix balance and diversity
Researchers have introduced a new training method for Generative Flow Networks (GFlowNets) called Rooted absorbed prefix Trajectory Balance (RapTB), designed to address issues like prefix collapse and length bias in lar…
-
New PACED-RL framework enhances LLM training efficiency
Researchers have proposed a new framework called PACED-RL that reinterprets the partition function in GFlowNets as a difficulty scheduler for LLM training. This approach leverages per-prompt expected reward signals, whi…
-
New method decomposes uncertainty in generative AI for scientific discovery
Researchers have developed a new method to decompose epistemic uncertainty in sequential generative models, particularly those used in AI-driven scientific discovery. By fitting polynomial chaos expansions to ensembles …
-
Stable GFlowNets algorithm improves training stability and fidelity
Researchers have introduced Stable GFlowNets, an algorithm designed to address training instability in Generative Flow Networks (GFlowNets). These networks are used for sampling states proportional to rewards but often …