RL on Qwen2-VL-2B with SAT dataset produces R1-like reasoning and 59.47% CVBench accuracy, outperforming base model by ~30% and SFT by ~2%.
Canonical reference
Title resolution pending
Canonical reference. 100% of citing Pith papers cite this work as background.
16
Pith papers citing it
Background
100% of classified citations
citation-role summary
background 6
citation-polarity summary
roles
background 6polarities
background 6representative citing papers
WE-MATH benchmark reveals most LMMs rely on rote memorization for visual math while GPT-4o has shifted toward knowledge generalization.
GAIA benchmark shows humans at 92% accuracy on simple real-world questions far outperform current AI systems at 15%, proposing this gap as a key milestone for general AI.
C-DSAC applies the Cramér distance to distributional value learning inside SAC and outperforms standard SAC on robotic benchmarks, with larger gains in complex environments due to confidence-driven conservative updates.
Template collapse is a distinct failure mode in agentic RL invisible to entropy; mutual information proxies diagnose it better and SNR-aware filtering using reward variance improves input-dependent reasoning and task performance across planning, math, navigation, and code tasks.
RLBFF extracts binary principles from human feedback to train reward models that outperform Bradley-Terry models on RM-Bench and JudgeBench and enable customizable inference-time focus for LLM alignment.
SkyReels-V2 produces infinite-length film videos via MLLM-based captioning, progressive pretraining, motion RL, and diffusion forcing with non-decreasing noise schedules.
FineWeb is a curated 15T-token web dataset that produces stronger LLMs than prior open collections, while its educational subset sharply improves performance on MMLU and ARC benchmarks.
REINFORCE-style variants outperform PPO, DPO, and RAFT in RLHF for LLMs by removing unnecessary PPO components and adapting the simpler method to LLM alignment characteristics.
DPOP is a new loss function that prevents DPO from lowering preferred response likelihoods and outperforms standard DPO on diverse datasets, MT-Bench, and enables Smaug-72B to exceed 80% on the Open LLM Leaderboard.
HAPO is a new token-level policy optimization method for LLMs that continuously adapts four optimization stages using entropy, claiming consistent gains over DAPO on math, code, and logic tasks.
InvDesFlow-AL combines active learning with diffusion generative models to improve crystal structure prediction accuracy by 33% and identifies Li2AuH6 as a candidate BCS superconductor with 140 K transition temperature.
Jailbreak prompts with adversarial suffixes have high GPT-2 perplexity, and a LightGBM model on perplexity and length detects most attacks.
Optimizing the noise schedule, preparing a balanced bucketed dataset, and aligning outputs with human preferences enables Playground v2.5 to reach state-of-the-art aesthetic quality across aspect ratios.
This survey discusses key components and challenges for Personal LLM Agents and reviews solutions for their capability, efficiency, and security.
A position and survey paper that identifies convergence between neuroscience, AGI, and neuromorphic computing and outlines four key integration challenges.
citing papers explorer
-
R1-Zero's "Aha Moment" in Visual Reasoning on a 2B Non-SFT Model
RL on Qwen2-VL-2B with SAT dataset produces R1-like reasoning and 59.47% CVBench accuracy, outperforming base model by ~30% and SFT by ~2%.
-
We-Math: Does Your Large Multimodal Model Achieve Human-like Mathematical Reasoning?
WE-MATH benchmark reveals most LMMs rely on rote memorization for visual math while GPT-4o has shifted toward knowledge generalization.
-
GAIA: a benchmark for General AI Assistants
GAIA benchmark shows humans at 92% accuracy on simple real-world questions far outperform current AI systems at 15%, proposing this gap as a key milestone for general AI.
-
Distributional Reinforcement Learning via the Cram\'er Distance
C-DSAC applies the Cramér distance to distributional value learning inside SAC and outperforms standard SAC on robotic benchmarks, with larger gains in complex environments due to confidence-driven conservative updates.
-
RAGEN-2: Reasoning Collapse in Agentic RL
Template collapse is a distinct failure mode in agentic RL invisible to entropy; mutual information proxies diagnose it better and SNR-aware filtering using reward variance improves input-dependent reasoning and task performance across planning, math, navigation, and code tasks.
-
RLBFF: Binary Flexible Feedback to bridge between Human Feedback & Verifiable Rewards
RLBFF extracts binary principles from human feedback to train reward models that outperform Bradley-Terry models on RM-Bench and JudgeBench and enable customizable inference-time focus for LLM alignment.
-
SkyReels-V2: Infinite-length Film Generative Model
SkyReels-V2 produces infinite-length film videos via MLLM-based captioning, progressive pretraining, motion RL, and diffusion forcing with non-decreasing noise schedules.
-
The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale
FineWeb is a curated 15T-token web dataset that produces stronger LLMs than prior open collections, while its educational subset sharply improves performance on MMLU and ARC benchmarks.
-
Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs
REINFORCE-style variants outperform PPO, DPO, and RAFT in RLHF for LLMs by removing unnecessary PPO components and adapting the simpler method to LLM alignment characteristics.
-
Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive
DPOP is a new loss function that prevents DPO from lowering preferred response likelihoods and outperforms standard DPO on diverse datasets, MT-Bench, and enables Smaug-72B to exceed 80% on the Open LLM Leaderboard.
-
Heterogeneous Adaptive Policy Optimization: Tailoring Optimization to Every Token's Nature
HAPO is a new token-level policy optimization method for LLMs that continuously adapts four optimization stages using entropy, claiming consistent gains over DAPO on math, code, and logic tasks.
-
InvDesFlow-AL: active learning-based workflow for inverse design of functional materials
InvDesFlow-AL combines active learning with diffusion generative models to improve crystal structure prediction accuracy by 33% and identifies Li2AuH6 as a candidate BCS superconductor with 140 K transition temperature.
-
Detecting Language Model Attacks with Perplexity
Jailbreak prompts with adversarial suffixes have high GPT-2 perplexity, and a LightGBM model on perplexity and length detects most attacks.
-
Playground v2.5: Three Insights towards Enhancing Aesthetic Quality in Text-to-Image Generation
Optimizing the noise schedule, preparing a balanced bucketed dataset, and aligning outputs with human preferences enables Playground v2.5 to reach state-of-the-art aesthetic quality across aspect ratios.
-
Personal LLM Agents: Insights and Survey about the Capability, Efficiency and Security
This survey discusses key components and challenges for Personal LLM Agents and reviews solutions for their capability, efficiency, and security.
-
Bridging Brains and Machines: A Unified Frontier in Neuroscience, Artificial Intelligence, and Neuromorphic Systems
A position and survey paper that identifies convergence between neuroscience, AGI, and neuromorphic computing and outlines four key integration challenges.