SA-GFN corrects biases in GFlowNets due to graph symmetries.
problem Systematic biases in state transition probability computations.
method Incorporates symmetry corrections into the learning process through reward scaling.
result Eliminates need for explicit state transition computations.
RBMLE learns from reward bias to optimize bandit decisions.
problem Optimizing decisions in stochastic multi-armed bandits.
method Reward-Biased Maximum Likelihood Estimation (RBMLE) for SMABs.
result RBMLE achieves optimal regret bounds for Gaussian and sub-Gaussian bandits.
New approach learns human planning algorithms for reward inference.
problem Learning reward functions from human demonstrations with biases.
method Data-driven approach to learn planning algorithms directly from demonstrations.
result Mixed results: better reward inference but at cost of differentiability.
NeuralRBMLE tackles explore-exploit trade-offs in contextual bandits with neural networks.
problem Stochastic contextual bandit problem with general bounded reward functions.
method Reward-biased maximum likelihood estimation with neural networks to enforce exploration.
result Both NeuralRBMLE variants achieve O ~ ( T ) \widetilde{\mathcal{O}}(\sqrt{T}) O ( T ) regret. The paper tackles bandit problems with biased offline data by using causal methods.
problem Improving bandit algorithms with biased offline data that includes confounding and selection biases.
method Formalizes the problem from a causal perspective, categorizes biases, and derives robust bounds for each arm.
result Causal bounds can guide the bandit agent to learn a nearly-optimal decision policy and consistently reduce asymptotic regret.
The paper proposes biased advantage estimates for reinforcement learning, improving efficiency and performance.
problem Estimating advantages in reinforcement learning algorithms.
method A family of estimates based on order statistics over the path ensemble.
result Biased estimates can significantly benefit reinforcement learning, especially in environments with sparse rewards or critical actions.
Mitigates biases in reward models using variational inference.
problem Spurious correlations in reward models that align large language models with human preferences.
method Formulates data-generating process, identifies non-spurious latent variables, and uses variational inference to recover them.
result Effective mitigation of spurious correlation issues, yielding more robust reward models.
Study incentivizes exploration in non-stationary MAB with compensation.
problem Incentivized exploration for non-stationary stochastic bandits with biased feedback.
method Proposed algorithms for abruptly-changing and continuously-changing non-stationary environments.
result Achieves sublinear regret and compensation over time.
New reward function improves GAIL performance in task-based environments.
problem Reward bias in adversarial imitation learning.
method Proposed a new reward function to overcome existing biases.
result New reward function outperforms existing methods in task-based environments.
Bayesian optimization improved for biased data.
problem Adversarial bias in observations, especially hidden confounders.
method Reduction to dueling bandits, information-directed sampling (IDS).
result First efficient kernelized algorithm with regret guarantees.
New algorithms improve linear bandit performance with low computation.
problem Optimizing reward in linear stochastic bandits.
method Reward-biased maximum likelihood method modified for linear and generalized linear bandits.
result New policies achieve order-optimality and competitive empirical performance.
Incentivized exploration improves MAB performance under reward drift.
problem Improving exploration in multi-armed bandits with biased feedback.
method Analysis of three MAB algorithms (UCB, ε-Greedy, Thompson Sampling) under drifted reward feedback.
result All algorithms achieve O ( log T ) \mathcal{O}(\log T) O ( log T ) regret and compensation under drifted reward. RAFT fine-tunes models using high-quality samples to align them with human preferences.
problem Aligning generative models with human ethics and preferences.
method RAFT selects high-quality samples, discards undesired behavior, and fine-tunes the model on filtered samples.
result RAFT improves model performance in reward learning and automated metrics.
CausalRM models rewards from user feedback, overcoming noise and bias.
problem Aligning language models with user preferences from noisy, biased feedback.
method Causal-theoretic reward modeling framework addressing noise and bias in observational feedback.
result CausalRM learns accurate reward signals from noisy and biased observational feedback.
This paper proposes a novel approach for constructing effective personalized policies when the observed data lacks counter-factual information, is biased and possesses many features. The approach is applicable in a wide variety of settings from healthcare to advertising to education to finance. These settings have in c…
The paper tackles fair sequential decision making with biased linear bandit feedback.
problem Fair sequential decision making with biased linear bandit feedback.
method Phased elimination algorithm to correct unfair evaluations, establishing upper bounds on regret.
result The worst-case regret is smaller than O ( κ ∗ 1 / 3 log ( T ) 1 / 3 T 2 / 3 ) \mathcal{O}(κ_*^{1/3}\log(T)^{1/3}T^{2/3}) O ( κ ∗ 1/3 log ( T ) 1/3 T 2/3 ) . Study improves unbiased recommender learning by addressing missing-reward bias.
problem Data bias caused by missing-reward observations in recommender systems.
method Proposes a novel estimator using propensity scores to mitigate both position and reward bias.
result The proposed estimator outperforms other methods, even with increased reward observation bias.
A new reinforcement learning framework for multi-reward processing.
problem Understanding and modeling multi-reward interactions in complex systems.
method Two-stream reward processing with biological associations.
result Agents can react differently to different types of rewards.
Local search improves GFlowNets' ability to generate high-reward samples.
problem GFlowNets struggle with over-exploration in high-reward space.
method Local search focusing on high-reward samples via backtracking and reconstruction.
result Significant performance improvement in biochemical tasks.
New policy tackles evolving externalities in contextual bandits.
problem Difficulty in recovering from wrong decisions over time.
method Rejection-based policy to achieve low regret.
result Low regret achieved regardless of reward matrix structure.
Recent studies have shown that reinforcement learning (RL) models are vulnerable in various noisy scenarios. For instance, the observed reward channel is often subject to noise in practice (e.g., when rewards are collected through sensors), and is therefore not credible. In addition, for applications such as robotics, …
A two-stream reinforcement learning model improves decision-making across human and neuropsychiatric studies.
problem Improving reinforcement learning models to better simulate human decision-making and neuropsychiatric conditions.
method Proposes a two-stream reinforcement learning model that processes positive and negative rewards and incorporates reward-processing biases.
result The two-stream model outperforms standard Q-learning and SARSA methods on various tasks and datasets.
Language models perform worse with implicit reward models than explicit ones.
problem Understanding why implicit reward models generalize worse than explicit ones.
method Investigated the root cause of the generalization gap between IM-RMs and EX-RMs.
result Implicit reward models rely more on superficial token-level cues, leading to worse generalization.
Active inference enhances RL by balancing exploration and exploitation.
problem Traditional RL's balance between exploration and exploitation is often suboptimal.
method Developed a new decision-making objective based on active inference.
result The new algorithm successfully balances exploration and exploitation on various RL benchmarks.
The paper tackles noisy multi-armed bandit problems with improved regret guarantees.
problem Tackling noisy evaluations in multi-armed bandit problems.
method Derives different algorithmic approaches and theoretical guarantees based on the type of observation functions.
result Improved regret guarantees for noisy linear functions of true rewards.
We identify two issues with the family of algorithms based on the Adversarial Imitation Learning framework. The first problem is implicit bias present in the reward functions used in these algorithms. While these biases might work well for some environments, they can also lead to sub-optimal behavior in others. Secondl…
New method uses simple sensor intentions to learn complex tasks.
problem Defining reward schemes for exploration in robotic systems.
method Introduce simple sensor intentions (SSIs) to define auxiliary tasks.
result Learning system can solve complex robotic tasks using only raw sensor streams.
RPPs improve deep learning models with soft equivariance constraints.
problem Balancing expressiveness and inductive biases in deep learning.
method Introducing Residual Pathway Priors (RPPs) to convert hard constraints into soft priors.
result RPPs enable models to learn structured solutions while retaining flexibility.
The paper tackles linear bandits with projections, achieving optimal regret.
problem Learning the best arm in a linear bandit model with unobservable projection rewards.
method Developed strategies for both finite and infinite arms, achieving optimal regret bounds.
result Achieved optimal regret bounds for both finite and infinite arms.
A RL framework selects features to balance bias and accuracy dynamically.
problem Bias in automated feature selection when predictors are correlated.
method Multi-component reward function with policy gradient for dynamic regularization and bias mitigation.
result Model balances fairness and accuracy during training.
Paper proposes DAC-ML, a cognitive architecture that learns quickly from few episodes.
problem Sample inefficiency in AI learning action policies.
method Incorporates hippocampus-inspired sequential memory system into DAC theory of mind.
result DAC-ML rapidly converges to effective action policies maximizing reward.
New algorithm balances user reward and statistical inference by mixing TS with UR based on difference size.
problem Combining statistical inference with user reward in adaptive experiments.
method TS-PostDiff algorithm that uses UR when differences are small and TS when large.
result TS-PostDiff reduces false positives and increases statistical power for small differences, while maximizing reward for large ones.
This work explains RL policies using causal models, revealing important patterns and failures.
problem Understanding why RL policies succeed or fail in complex, high-dimensional systems.
method Developed a nonlinear Causal Model Reduction framework to learn simplified causal models from RL policy actions and rewards.
result The approach can uncover important behavioral patterns and failure modes in trained RL policies.
A new algorithm for fair decision-making in bandit problems with biased feedback.
problem Fair allocation of resources across sensitive groups in bandit problems with biased feedback.
method Proposes a novel algorithm that learns a societal bias term to mitigate unfairness in multi-armed bandit problems.
result The proposed algorithm achieves a theoretical bound on regret and demonstrates fairness in synthetic and real-world data.
New algorithm mitigates affinity bias in hiring feedback loops.
problem Mitigating affinity bias in hiring decisions to avoid unconscious favoritism.
method Introducing affinity bandits, a new bandit variant that accounts for evolving biased feedback.
result Elimination-style algorithm nearly matches the derived regret bound, outperforming classical algorithms.
New RL algorithm tackles adversarial RMAB with unknown transitions and bandit feedback.
problem Learning in episodic RMAB with unknown transition functions and adversarial rewards.
method Developed a novel RL algorithm with a biased reward estimator and an index policy.
result Achieved i l d e O ( H T ) ilde{\mathcal{O}}(H\sqrt{T}) i l d e O ( H T ) regret bound for adversarial RMAB. DTS improves inference-time alignment of diffusion models with less compute.
problem Inference-time alignment of diffusion models suffers from inaccurate value estimation and inefficient reuse of past computations.
method Diffusion Tree Sampling (DTS) uses a tree-based approach to propagate terminal rewards and iteratively refine value estimates.
result DTS produces asymptotically exact samples and matches the FID of best-performing baselines with up to 10x less compute.
This work uses statistical bootstrapping to provide accurate confidence intervals for policy value in reinforcement learning.
problem Bias in estimating policy value using empirical transitions and rewards.
method Statistical bootstrapping to produce calibrated confidence intervals for the true policy value.
result Statistical bootstrapping can yield correct confidence intervals under certain conditions, and mechanisms are proposed to mitigate these conditions.
Develops methods to correct bias in AI feedback for more accurate alignment.
problem Systematic bias in AI feedback compared to human labels.
method Two debiased alignment methods: DDPO and DIPO.
result Methods improve alignment efficiency and performance close to human-labeled data.
New meta-learning approach for bandit policies that achieve high average reward.
problem Designing bandit policies that balance between worst-case and Bayesian assumptions.
method Differentiable parameterized policies optimized using policy gradients.
result Proposed algorithm achieves low regret and is practical for various bandit problems.
New metrics improve scRNA-seq perturbation modeling by reducing mode collapse.
problem Outperformed by simple mean prediction in scRNA-seq perturbation modeling.
method Introduce DEG-aware metrics (WMSE, R w 2 ( Δ ) R^{2}_{w}(Δ) R w 2 ( Δ ) ) and negative/positive baselines. result WMSE loss function reduces mode collapse and improves model performance.
HIGhER uses language to generate new instructions for better learning from mistakes.
problem Improving instruction following in reinforcement learning environments.
method Hindsight Generation for Experience Replay (HIGhER) that learns from mistakes and relabels episodes.
result HIGhER enhances instruction following in reinforcement learning environments.
Recently, Reinforcement Learning (RL) approaches have demonstrated advanced performance in image captioning by directly optimizing the metric used for testing. However, this shaped reward introduces learning biases, which reduces the readability of generated text. In addition, the large sample space makes training unst…
The paper shows how curated synthetic data can optimize human preferences in generative models.
problem Contamination of web-scale datasets by synthetic data affects future model training.
method Theoretical study of iterated retraining of generative models with curated synthetic data.
result Data curation can be seen as an implicit preference optimization mechanism, maximizing expected reward.
Agent learns directed exploration policies to improve performance in hard games.
problem Improving exploration in complex games.
method Episodic memory-based intrinsic reward, self-supervised inverse dynamics, UVFA framework.
result Doubles performance in hard exploration games, achieves non-zero rewards in Pitfall!.
HRL improves open-domain dialog models by optimizing long-term conversational goals.
problem Challenges in open-domain dialog generation, including repetitive outputs, difficulty tracking conversational goals, and inappropriate text.
method Proposes VHRL, a hierarchical reinforcement learning approach using policy gradients to tune utterance-level embeddings of a variational sequence model.
result Significant improvements in human evaluation and automatic metrics over state-of-the-art dialog models.
Existing model-based reinforcement learning methods often study perception modeling and decision making separately. We introduce joint Perception and Control as Inference (PCI), a general framework to combine perception and control for partially observable environments through Bayesian inference. Based on the fact that…
A method to reduce bias in model-based policy evaluation by shifting operators.
problem Bias in value function computation from noisy estimated models.
method Operator shifting method to reduce the residual norm error.
result The shifting factor is always positive and upper bounded by $1+O\left(1/n
ight)$ .