Tagged articles
curriculum learning
20 articles · Page 1 of 1
The article analyses how self‑evolving AI agents shift from simple tool use to autonomous planning, proposes a fast‑slow thinking architecture, curriculum learning, organizational structures, continuous evaluation, and a four‑layer safety framework to ensure they grow responsibly while collaborating with humans.
AI agentsAgent SafetyContinuous Evaluation
0 likes · 17 min read

The paper reveals that visual language models often rely on chronological bias, mistaking later frames for progress, and introduces EgoTSR—a 46‑million‑sample ego‑centric dataset and three‑stage curriculum that teaches models to assess task state, evaluate with forward‑reverse tests, and achieve over 92% accuracy on long‑term robotic tasks.
Roboticschronological-biascurriculum learning
0 likes · 11 min read

GenShield, a unified autoregressive framework introduced by researchers from Peking University, combines explainable AI‑generated image detection with controllable artifact correction, leveraging a two‑stage Visual Chain‑of‑Thought curriculum and the newly built GenShield‑Set dataset to achieve state‑of‑the‑art performance on both detection and repair benchmarks.
AI-generated image detectionGenShieldartifact correction
0 likes · 8 min read

The article showcases five ICML 2026 papers from the Taotian Group that tackle core multimodal AI challenges—interactive video try‑on, high‑resolution vision, e‑commerce video reasoning, sparse‑reward reinforcement learning, and curriculum learning for large language models—detailing their problem statements, novel solutions, and strong experimental results.
ICML 2026Reinforcement Learningbenchmark
0 likes · 15 min read

The paper introduces PACE, a Parameter‑Change based Unsupervised Environment Design method that evaluates training levels by the magnitude of induced policy‑parameter updates, offering a low‑variance, computationally cheap signal that consistently outperforms prior UED approaches on MiniGrid and Craftax benchmarks.
CraftaxICML 2026MiniGrid
0 likes · 11 min read

The paper introduces PACE, a Parameter Change Environment Design method that evaluates training levels by measuring induced policy parameter updates, offering a low‑variance learning‑progress signal that outperforms prior UED approaches on MiniGrid and Craftax benchmarks, achieving higher success rates and more stable generalization.
CraftaxICML 2026MiniGrid
0 likes · 10 min read

This article reviews a recent paper that proposes a drift‑aware data‑stream system integrating machine‑learning‑based adaptive control into financial data management, introducing a parametric data‑operation module, a gradient‑based bi‑level optimizer, and a curriculum planner to improve model robustness and risk‑adjusted returns in non‑stationary markets.
Data AugmentationQuantitative Financeadaptive data synthesis
0 likes · 18 min read

The article details how AliExpress leverages a multimodal large‑model pipeline to automatically fill missing product attributes, discover schema gaps, and generate rich application tags, describing the technical challenges, training stages, experimental results, and business impact of the solution.
chain-of-thoughtcurriculum learninge‑commerce AI
0 likes · 28 min read

RAVEN is a reinforcement‑reasoning framework that combines curriculum learning with hierarchical rewards to enable multimodal large language models to accurately locate and classify violation segments in advertisement videos, even under noisy, large‑scale industrial data.
AdvertisingReinforcement Learningcurriculum learning
0 likes · 17 min read


Amap TechNov 4, 2025 · Artificial Intelligence
This article introduces Spacetime‑GR, a large‑scale generative recommendation model that integrates hierarchical geographic POI indexing and spatiotemporal token encoding to enhance POI prediction for Amap, detailing its pre‑training pipeline, data cleaning, curriculum learning strategy, experimental results, scaling law observations, and the resulting improvements in hit rate and discovery rate.
AmapGenerative AIPOI recommendation
0 likes · 14 min read

SEAgent introduces a self‑evolving framework that enables a GUI agent to master unfamiliar software through autonomous exploration and experience learning, leveraging a curriculum generator, a world‑state model, and GRPO‑based reinforcement with adversarial imitation, achieving state‑of‑the‑art performance on OSWorld.
Experience LearningGUI automationReinforcement Learning
0 likes · 6 min read


ArchitectMar 9, 2025 · Artificial Intelligence
The author reports a series of reinforcement‑learning‑based fine‑tuning experiments on a 0.5‑billion‑parameter Qwen‑0.5VB instruct model using the KK dataset, detailing reward design adjustments, curriculum‑style data scaling, observed convergence issues, and hypotheses about why small models fail to develop long reasoning chains.
LLM fine-tuningReinforcement LearningReward Design
0 likes · 11 min read

The author experiments with reinforcement‑learning‑from‑human‑feedback on a 0.5B Qwen instruct model using Logic‑RL and Open‑R1, discovers that reward mis‑design and curriculum learning cause the model to produce overly short or incorrect reasoning chains on knight‑and‑knave puzzles, and analyses the underlying causes.
Logic ReasoningRLHFReward Design
0 likes · 11 min read

The paper introduces TAPIR, a task‑aware curriculum planning framework that distills instruction‑following abilities from black‑box LLM teachers into smaller student models by filtering difficult prompts, resampling tasks, enhancing response styles, and iteratively optimizing across multiple training rounds, achieving superior performance on benchmark evaluations.
LLM distillationTAPIRcurriculum learning
0 likes · 10 min read

This comprehensive guide walks you through every stage of LLM pretraining—from data sourcing, cleaning, and deduplication to tokenizer design, model architecture choices, training framework selection, optimization tricks, and evaluation methods—highlighting common pitfalls and practical solutions for building robust models.
LLM Pretrainingcurriculum learningdata cleaning
0 likes · 34 min read

This talk outlines the fundamental challenges of multi‑agent decision large models, introduces three core design priors—action semantic networks, permutation invariance/equivariance, and cross‑task automated curriculum learning— and demonstrates how these concepts improve performance across diverse environments such as StarCraft, Neural‑MMO, and SMAC.
AIMulti-Agent Reinforcement Learningaction semantics
0 likes · 12 min read

The paper presents a Curriculum‑Guided Bayesian Reinforcement Learning (CBRL) framework that models ROI‑constrained real‑time bidding as a partially observable constrained MDP, using hard‑margin indicator rewards and a curriculum of relaxed proxy problems to achieve fast, constraint‑satisfying, Bayes‑optimal policies that outperform existing methods on large‑scale industrial data.
Bayesian RLMDPROI constraint
0 likes · 15 min read

This article explains how treating negative response candidates with varying difficulty through a hierarchical curriculum learning framework—combining corpus‑level and instance‑level curricula—enhances dialogue response selection models, backed by experiments on Douban, Ubuntu, and E‑Commerce datasets.
curriculum learningdialogue response selectionhierarchical learning
0 likes · 8 min read

The paper introduces a hybrid curriculum learning framework that tackles emotion shift and confusing labels in emotion recognition in conversation by applying nested curriculum stages at both conversation and utterance levels, enabling a progressive easy‑to‑hard training that markedly boosts classic ERC model performance across four public datasets and is already deployed in Alibaba’s entertainment AI brain script health‑check service.
Emotion Recognitionconversation analysiscurriculum learning
0 likes · 2 min read

This article reviews the challenges of model‑free reinforcement learning, especially sparse reward issues exemplified by Montezuma’s Revenge, and surveys recent approaches such as expert demonstrations, curriculum learning, self‑play, hierarchical reinforcement learning, and count‑based exploration to mitigate these problems.
Model-freeSelf-Playcurriculum learning
0 likes · 12 min read
