QVal is a new evaluation framework that directly measures dense supervision quality via Q-alignment to a reference policy, showing simple prompting baselines outperform 21 other methods across environments and models.
hub
RL-VLM-F: Reinforcement learn- ing from vision language foundation model feedback
15 Pith papers cite this work, alongside 6 external citations. Polarity classification is still indexing.
15
Pith papers citing it
6
external citations · Pith
abstract
Reward engineering has long been a challenge in Reinforcement Learning (RL) research, as it often requires extensive human effort and iterative processes of trial-and-error to design effective reward functions. In this paper, we propose RL-VLM-F, a method that automatically generates reward functions for agents to learn new tasks, using only a text description of the task goal and the agent's visual observations, by leveraging feedbacks from vision language foundation models (VLMs). The key to our approach is to query these models to give preferences over pairs of the agent's image observations based on the text description of the task goal, and then learn a reward function from the preference labels, rather than directly prompting these models to output a raw reward score, which can be noisy and inconsistent. We demonstrate that RL-VLM-F successfully produces effective rewards and policies across various domains - including classic control, as well as manipulation of rigid, articulated, and deformable objects - without the need for human supervision, outperforming prior methods that use large pretrained models for reward generation under the same assumptions. Videos can be found on our project website: https://rlvlmf2024.github.io/
hub tools
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
Self-distillation from a caption-conditioned video diffusion model to an image-and-prompt-conditioned executor, enhanced by RL from VLM feedback, enables task solving in world models.
CV-Arena is a new 12K-pair benchmark for instruction-guided real-image editing with 16 task types, CogRetriever curation, and Active Elo mixed human-AI evaluation that finds gaps in 21 models and presents CV-Agent.
Formalizes Reasoning Portability (RP) and proposes RDB-CL to modulate per-sample KL regularization in RLVR for MLLM continual learning, achieving +12.0% Last accuracy over vanilla RLVR baseline by preserving reusable reasoning on high-RP samples.
FPL trains a language-conditioned reward model from per-axis human preferences and a reward-conditioned policy, reporting 38-point average success gains over sparse-reward and binary-preference baselines on six manipulation tasks.
MAPL trains quadruped locomotion policies from LLM-generated multi-objective trajectory preferences and matches or exceeds expert-designed reward performance in four environments without manual reward engineering.
Success Visitation Matching uses a discriminator to turn sparse outcome rewards into dense process rewards by matching visitations of successful episodes, provably preserving the optimal policy and speeding up robotic RL finetuning.
A framework learns invariant symbolic reward functions from few demonstrations that generalize zero-shot to variations in robotic manipulation tasks.
Expecting over scoring-token logits yields continuous, scalable verification that improves agent trajectory selection and dense RL rewards across coding, robotics, and medical benchmarks.
MAGNIFIED applies RL fine-tuning to MLLMs for autonomous driving motion planning, yielding over 10.5% lower overlap rate and 38.9% lower off-road rate than SFT baseline on Waymo Open Motion Dataset.
AtlasVA organizes VLM agent memory into spatial heatmaps, visual exemplars, and symbolic skills, evolving atlases from trajectories to act as potential-based shaping rewards in teacher-free reinforcement learning.
Reflective Self-Adaptation combines failure-reflective reinforcement learning with success-guided imitation learning to enable faster and more reliable task adaptation for pre-trained Vision-Language-Action models.
UAV-VL-R1 combines SFT and multi-stage GRPO reinforcement learning on a new 50,019-sample HRVQA-VL dataset to deliver substantially higher zero-shot accuracy on UAV visual reasoning tasks than both its 2B baseline and a 72B-scale model.
Presents a hybrid agentic framework using MediaPipe, Llama-4-scout VLM, LangGraph orchestration, and RAG for holistic athlete profiling aligned with SAI protocols.
A survey provides a task-based formalization of meta-learning and meta-RL while chronicling algorithms that lead to DeepMind's Adaptive Agent.
citing papers explorer
-
QVal: Cheaply Evaluating Dense Supervision Signals for Long-Horizon LLM Agents
QVal is a new evaluation framework that directly measures dense supervision quality via Q-alignment to a reference policy, showing simple prompting baselines outperform 21 other methods across environments and models.
-
World Model Self-Distillation: Training World Models to Solve General Tasks
Self-distillation from a caption-conditioned video diffusion model to an image-and-prompt-conditioned executor, enhanced by RL from VLM feedback, enables task solving in world models.
-
CV-Arena: An Open Benchmark for Instructional Computer Vision Problem Solving with Human-AI Collaborative Preferences
CV-Arena is a new 12K-pair benchmark for instruction-guided real-image editing with 16 task types, CogRetriever curation, and Active Elo mixed human-AI evaluation that finds gaps in 21 models and presents CV-Agent.
-
Reasoning Portability: Guiding Continual Learning for MLLMs in the RLVR Era
Formalizes Reasoning Portability (RP) and proposes RDB-CL to modulate per-sample KL regularization in RLVR for MLLM continual learning, achieving +12.0% Last accuracy over vanilla RLVR baseline by preserving reusable reasoning on high-RP samples.
-
Freeform Preference Learning for Robotic Manipulation
FPL trains a language-conditioned reward model from per-axis human preferences and a reward-conditioned policy, reporting 38-point average success gains over sparse-reward and binary-preference baselines on six manipulation tasks.
-
MAPL: Multi-Objective Preference Learning for Robot Locomotion
MAPL trains quadruped locomotion policies from LLM-generated multi-objective trajectory preferences and matches or exceeds expert-designed reward performance in four environments without manual reward engineering.
-
Learning Process Rewards via Success Visitation Matching for Efficient RL
Success Visitation Matching uses a discriminator to turn sparse outcome rewards into dense process rewards by matching visitations of successful episodes, provably preserving the optimal policy and speeding up robotic RL finetuning.
-
Beyond Pixels: Learning Invariant Rewards for Real-World Robotics From a Few Demonstrations
A framework learns invariant symbolic reward functions from few demonstrations that generalize zero-shot to variations in robotic manipulation tasks.
-
LLM-as-a-Verifier: A General-Purpose Verification Framework
Expecting over scoring-token logits yields continuous, scalable verification that improves agent trajectory selection and dense RL rewards across coding, robotics, and medical benchmarks.
-
MAGNIFIED: RL Fine-tuning of Multimodal Large Language Models for Motion Planning
MAGNIFIED applies RL fine-tuning to MLLMs for autonomous driving motion planning, yielding over 10.5% lower overlap rate and 38.9% lower off-road rate than SFT baseline on Waymo Open Motion Dataset.
-
AtlasVA: Self-Evolving Visual Skill Memory for Teacher-Free VLM Agents
AtlasVA organizes VLM agent memory into spatial heatmaps, visual exemplars, and symbolic skills, evolving atlases from trajectories to act as potential-based shaping rewards in teacher-free reinforcement learning.
-
Reflection-Based Task Adaptation for Self-Improving VLA
Reflective Self-Adaptation combines failure-reflective reinforcement learning with success-guided imitation learning to enable faster and more reliable task adaptation for pre-trained Vision-Language-Action models.
-
UAV-VL-R1: Generalizing Vision-Language Models via Supervised Fine-Tuning and Multi-Stage GRPO for UAV Visual Reasoning
UAV-VL-R1 combines SFT and multi-stage GRPO reinforcement learning on a new 50,019-sample HRVQA-VL dataset to deliver substantially higher zero-shot accuracy on UAV visual reasoning tasks than both its 2B baseline and a 72B-scale model.
-
Digitizing Coaching Intelligence: An Agentic Framework for Holistic Athlete Profiling using VLM and RAG
Presents a hybrid agentic framework using MediaPipe, Llama-4-scout VLM, LangGraph orchestration, and RAG for holistic athlete profiling aligned with SAI protocols.
-
Meta-Learning and Meta-Reinforcement Learning -- Tracing the Path towards DeepMind's Adaptive Agent
A survey provides a task-based formalization of meta-learning and meta-RL while chronicling algorithms that lead to DeepMind's Adaptive Agent.