Temporal Credit Assignment

Reinforcement learning enables large language model (LLM) agents to learn multi-step behaviors through interaction with their environments. However, rewards in many interactive tasks reflect only the final outcome, providing limited guidance on which intermediate decisions advance the task. Successful training trajectories contain intermediate states that can provide supervision for subsequent interactions. We introduce Trajectory-to-Step Policy Optimization (T2SPO), a method that uses past interaction trajectories to provide step-level feedback for policy learning. T2SPO derives remaining-distance targets from successful trajectories and pairs them with representations of the states visited along the way. Conditioned on these examples, a pretrained TabPFN regressor estimates the remaining distance to success at each state of a new rollout. Changes in this distance estimate across consecutive states yield auxiliary credit for agent steps alongside task-level supervision. As training proceeds, newly completed trajectories refresh the estimator's context, incorporating new experience without updating its parameters. Experiments with 1.5B and 7B language models on ALFWorld and WebShop show that T2SPO consistently improves overall task success over GRPO.

Process reward models (PRMs) have demonstrated notable effectiveness in test-time scaling and reinforcement learning by providing fine-grained signals for evaluating intermediate reasoning states, but their training relies heavily on costly process annotations. A natural way to alleviate this dependence is to complement limited process supervision with scalable outcome supervision. However, existing PRMs often model reasoning prefixes independently, providing no explicit mechanism for effectively using final outcome to guide the learning of intermediate reasoning states. We introduce Reasoning State Propagation (RSP), which represents each reasoning prefix with a binary validity state and models transitions between successive states across the reasoning trajectory. Specifically, RSP predicts a break probability that a valid state becomes invalid and a repair probability that an invalid state returns to valid. By propagating these transitions, RSP connects intermediate states to the final state, allowing process annotations to supervise intermediate states while outcome labels supervise the final state and can provide learning signals to preceding steps. Across reasoning search, response selection, and reinforcement learning, RSP consistently outperforms representative PRM baselines, with average improvements over Qwen2.5-Math-PRM of 5.6% in beam search and 2.1% in reinforcement learning.

Group-based reinforcement learning such as GRPO trains LLM agents by comparing rollouts sampled for each task, without a learned critic. In long-horizon settings, these rollouts revisit shared anchor states, offering cross-rollout evidence for step-level credit. Ideally, step-level credit should incorporate evidence beyond the realized suffixes observed at an anchor while aggregating alternative continuations according to their empirical frequencies. Visit-local averaging pools realized suffix returns at shared anchors and respects observed frequencies, but does not recursively propagate evidence across rollouts, whereas shortest-path estimators have global reach but allow a rarely observed route to dominate an anchor's value. We introduce Cross-Rollout Bellman Closure (CRBC), which merges each rollout group into a finite empirical process with absorbing success and failure boundaries and evaluates its behavior-policy Bellman fixed point with one linear solve. This fixed point uses the same empirical action and transition frequencies to propagate evidence through shared anchors and aggregate alternative continuations. Backing up the resulting state values through observed transitions yields action values, whose gain over the corresponding state value provides step-level credit. A corresponding finite-depth family recovers visit-local return averaging at zero depth and converges to the exact closure as depth increases. The normalized closure credit is combined with the trajectory-level group advantage for policy optimization, without additional environment rollouts. Across ALFWorld, WebShop, and Sokoban benchmarks with multiple model scales, CRBC consistently improves final performance and learning efficiency. For example, CRBC outperforms the strongest evaluated baseline by 5.59 percentage points on ALFWorld with Qwen2.5-1.5B-Instruct.

Sequential RecSys are central to modern personalization, exploiting user's historical interaction sequences to drive next-step decisions. Deep learning models, particularly CNN and Transformer-based architectures, have proven highly effective at capturing temporal dependencies in these histories. For transparency and trust, understanding which past interactions drive a given recommendation is increasingly important --- both for developers auditing model behavior and for users seeking a rationale. However, the non-linearities that give these models their predictive power also render them black boxes, making it difficult to attribute decisions to specific interactions. While gradient-based, perturbation-based, and attention-based explainability methods exist, a systematic benchmark of their faithfulness for sequential recommendation is missing. We address this gap by introducing a dual-model masking metric in which one model supplies per-timestep attribution scores and a separately trained, masking-robust probe measures the resulting change in predicted probability. Using this metric, we benchmark ten XAI methods across CNN, Transformer, SASRec, and BERT4Rec backbones on KuaiRand and MovieLens, complemented by analyses of temporal attribution patterns, item popularity confounding, and robustness to input corruption. Our key findings are: (1) gradient-based methods, particularly GradientSHAP and Integrated Gradients, yield the most faithful and robust attributions; (2) raw attention weights are unreliable, but gradient-weighted attention restores faithfulness on shorter sequences, with degradation on longer horizons as softmax attention probabilities converge toward uniform importance scores, diminishing the method's ability to identify informative interactions; and (3) temporal attribution patterns in faithful methods reflect genuine task structure rather than recency or popularity bias.

On-policy distillation (OPD) has emerged as an effective approach for large language model post-training, yet existing objectives face a trade-off between objective fidelity and optimization stability. Token-level OPD provides stable but local supervision, whereas sequence-level OPD captures future credit at the cost of horizon-dependent variance. We establish a unified temporal-credit view of these formulations, showing that practical token-level OPD can be interpreted as a temporal approximation to the sequence-level reverse-KL gradient. Building on this connection, we propose $γ$OPD, which uses discounted temporal credit assignment to balance long-horizon supervision and optimization stability, while admitting a horizon-independent variance bound. We further develop a reward-compatible bounded mixing (RBM) mechanism for $γ\mathrm{OPD}$ that balances verifiable outcome feedback with the discounted OPD advantage to move beyond purely teacher-dependent optimization. Experiments on mathematical and code reasoning demonstrate consistent improvements over existing OPD methods across vanilla, size-mismatched, and multi-teacher distillation settings.

In autoregressive forecasting, long prediction rollouts provide distant supervision, but backpropagation through time (BPTT) carries gradients from those losses through many autoregressive steps. Repeated Jacobian products can make distant gradients dominate the update while amplifying predictable signal and unpredictable innovation together; a large distant gradient therefore need not carry reliable learning signal. Motivated by this, we introduce Internal Dual-Wiener routing (Internal-DW), a backward-only intervention that preserves the full forward rollout and all step losses while reliability-weighting internal gradient routes. At each residual block, we derive bounded Wiener gains for the identity and nonlinear routes that balance preserving predictable learning signal against suppressing unpredictable variation, and estimate them from route-level gradient statistics and an explicit noise model. In a controlled system with known gradient signal-to-noise ratio (SNR), we show that distant gradients can grow even as their SNR falls, and that Internal-DW reduces error in recovering predictable gradient signals and improves forecasting. On four history-dominated, weak-drive testbeds, Internal-DW reduces forecast error by 5.2%-13.8% relative to full BPTT, outperforms gradient clipping and Jacobian regularization on three testbeds, with similar performance on shear flow, and outperforms validation-selected truncated BPTT (TBPTT) on three. It also extends or preserves the fitted optimal training-horizon range across these four testbeds. Across benchmarks, the current Internal-DW estimator has a clear applicability boundary: its benefit diminishes or reverses when usable history is limited or when the selected sampler fails to represent dominant drive-dependent variation. The results show that retaining long-horizon supervision does not require trusting every backward contribution equally.

Recurrent models provide a natural path to long-context modeling, yet models trained with backpropagation through time (BPTT) often fail beyond their training horizon. Classical analyses emphasize gradients that vanish or explode along temporal paths. However, dense per-token losses can still train a shared recurrent rule despite severe decay, showing that decay alone does not determine whether learning fails. We instead study state credit: the signal through which future losses reach earlier recurrent states before contributing to parameter updates. Accordingly, we intervene directly on state credit and propose Credit Stabilization through Time (CST). During backward propagation, CST locally rescales the state-credit signal to stabilize its norm without rotating the component being corrected, while leaving the forward computation unchanged. Because controlled synthetic tasks and real data exhibit different credit dynamics, we specialize CST to each regime. In both settings, CST improves performance beyond the training horizon, with gains observed at up to 128x the training length.

Sparse outcome feedback limits what robots can learn from unsuccessful attempts at complex manipulation. Failed multi-stage surgical attempts can contain grasps, lifts, or transfers worth reusing. In sparse-reward reinforcement learning, terminal rewards collapse such attempts to the same outcome, while scalar vision-language model (VLM) ratings reveal neither what progress merits credit nor when it occurred. We introduce phase-and-first-arrival feedback: one VLM query per recorded episode identifies the furthest visually verified task phase and when that phase is first reached, allowing the learner to reuse partial behavior and localize credit. We instantiate it in SurgPhaseBench, a phase-structured suite spanning rigid and deformable tasks, and evaluate it in simulation and hardware. Across five simulated tasks, our method reaches 75.2% mean success, compared with 52.1% for a reward based on Contrastive Language-Image Pre-training (CLIP) using the same visual input; the advantage persists when only the feedback representation changes. On hardware, the same record supports autonomous block picking and slip recovery. Together, these results show that trajectory-level visual supervision can preserve partial progress while providing the temporal credit needed for sparse-reward control.

Training multi-turn agentic workflows with reinforcement learning (RL) enables large language models to perform complex reasoning, use external tools, and conduct iterative search beyond single-turn settings. Yet multi-turn RL training remains highly unstable, often causing severe performance degradation as the number of turns increases. Through theoretical analysis, we identify three tightly coupled sources of instability: rollout-training context mismatch, weak turn-level credit assignment under sparse terminal rewards, and asynchronous policy drift when short and long trajectories are optimized under different policy versions. We show that these issues share a common structural origin in flattened trajectory optimization and address them through a unified reverse-turn formulation. We propose Reverse-Turn Policy Optimization (RTPO), which organizes multi-turn rollouts as sparse reverse trees and performs turn-level policy updates in temporal reverse order, aligning each decision with its downstream continuation. RTPO enables causally consistent turn-level credit assignment and on-policy continuation to control asynchronous drift. We provide theoretical guarantees showing that RTPO eliminates context mismatch and asynchronous drift under the proposed turn-level formulation, reduces credit bias, and converges to recursive optimality. Experiments on multi-turn agentic RL benchmarks show that RTPO improves upon trajectory- and turn-level baselines by 21.50% and 10.76%, respectively, highlighting its potential to support more stable training for tool-using agents.

Outcome-driven reinforcement learning offers a scalable way to post-train vision-language-action (VLA) policies from sparse task-success feedback. In common GRPO-based VLA post-training, one rollout-level advantage is applied to every action in the trajectory. A rollout that completes several valid stages but fails later can therefore penalize the actions that produced its earlier progress. We call this trajectory-level credit aliasing. Temporal GRPO addresses this problem by constructing detectable task stages, aligning each rollout with stage-specific action intervals, and comparing only rollouts that have entered the same stage. The resulting stage advantages are applied to their corresponding intervals in a single policy update. On RoboTwin 2.0, Temporal GRPO improves task success and sample efficiency, with consistent gains across task horizons. Controlled updates on LIBERO-Long preserve shared prerequisite stages and concentrate improvement at the first stage where rollout outcomes diverge.

Diffusion models have strong generative capabilities. However, their maximum likelihood training objective only focuses on reconstructing the data distribution, making it difficult to align with specific preferences. Reinforcement learning (RL) for preference alignment in diffusion models is promising but limited by reward sparsity. Since a single reward cannot support optimization, existing RL methods usually backpropagate the final reward to all previous steps. However, denoising is stage-wise, with distinct semantics and controllability. Repeating the final reward across all steps creates a temporal objective mismatch, encouraging reward shortcuts that lead to reward hacking. At the same time, due to reward backfilling, each time step receives the same reward, making it impossible to distinguish between actions, thereby weakening the optimization process. To resolve this issue, we propose Stage-Guided Per-Step Optimization (SGPO) for diffusion models, which jointly leverages signal-to-noise ratio and semantic changes to identify generation stages and adaptively assign stage-specific objectives. Early denoising is chaotic and far from the final reward, resulting in weak reward-behavior correlation. This stage should prioritize exiting the chaotic state. In the mid stage, the latent transitions to a stable structure, where the final reward better corresponds to generative behavior. Therefore, this stage optimizes the final reward while exploring diversity to avoid early convergence to a single mode. In the late stage, the latent's core structure is largely fixed, and preference optimization mainly amplifies local details, risking overfitting. Therefore, stable convergence is preferred to avoid quality degradation. Results from 16 comparative experiments validate SGPO. Our method achieves 26.7% average gains in generative quality and 36.7% higher convergence speed.

Recent advances in preference alignment for diffusion-based video generation, particularly via Direct Preference Optimization (DPO), have significantly improved visual quality. However, temporally sparse artifacts such as motion collapse, object flickering, and color oversaturation remain a major barrier to perceptual realism. Existing methods struggle with these issues due to two key limitations: (1) the preference attribution bottleneck, where offline human annotations are costly and fail to accurately capture learning dynamics, while online reward signals are rollout-aware but often unstable and biased; and (2) temporal credit misallocation, where uniformly applied supervision cannot effectively target the brief segments in which artifacts occur. To address these challenges, we propose concentrated Implicit Preference Optimization (cIPO), a post-training framework for video diffusion models. cIPO derives implicit preference signals directly from the denoising process: given a real video, the model adds forward noise and reconstructs it via iterative denoising, treating the original as the preferred sample and the reconstruction as the dispreferred one. This formulation captures inference-time errors without requiring human annotations or external reward models. Moreover, frame-level discrepancies between original and reconstructed videos reveal when failures occur. cIPO leverages this by computing temporal reconstruction errors and concentrating optimization on high-error segments, enabling more precise correction of failure-prone regions. Extensive experiments demonstrate that cIPO consistently enhances video authenticity and temporal coherence across multiple datasets, highlighting the effectiveness and efficiency of implicit preference with temporally concentrated optimization.

The Credit Assignment Problem (CAP) is fundamental to developing efficient and explainable Reinforcement Learning (RL) agents. Existing frameworks, whether relying on temporal contiguity or hindsight-conditioned reward reweighting, frequently fail to attribute properly between an agent's policy (skill) and environmental stochasticity (luck). A principled approach to CAP must isolate the true causal drivers of observed outcomes from spurious correlations and environmental randomness. We introduce Counterfactual Shapley Credit Assignment, a novel framework grounded in causal theory that attributes credit and blame via the Counterfactual Shapley Value ($φ$-value). By redistributing environmental rewards, $φ$-values enhance temporal credit assignment across three critical dimensions: sparse causality, high stochasticity, and delayed rewards, all while preserving the optimal policy. We derive a consistent estimator that computes $φ$-values efficiently, enabling a new class of policy gradient methods, $φ$-PPO, combined with Prioritized Trajectory Replay (PTR). Empirical results demonstrate that $φ$-values align precisely to the ground truth causes of task rewards with superior sample efficiency in challenging environments where prior state-of-the-art methods fail to converge.

Learned dynamics models often answer global physical questions, such as fault severity or impact stiffness, by pooling a per-step feature sequence into one readout vector. This sequence-to-global interface creates an under-studied temporal credit problem: with only trajectory-level supervision, a model can predict accurately in training conditions while reading from abundant smooth correlates rather than the brief physical events that determine the target. We call this failure temporal credit dilution. It is not exposed by the training loss and is not removed by standard physics-informed residuals, because the error lies in where the global readout assigns functional credit. We introduce Credit-in-Event, an interface-level probe for measuring how much pooled credit lands on event steps, and prove in closed form that a pooled linear reader routes credit to a spurious background channel as the event fraction shrinks. We then propose CREST, a training-free and label-free readout that estimates a transient event core from learned features and re-anchors the pooled representation through event-versus-rest contrast. Across simulated gear and impact systems, recurrent and attention encoders, and public bearing vibration data, CREST reduces out-of-distribution error while restoring event credit. Ablations show that stable-step selection and receptive-field shrinking fail, confirming that the gain comes from event-core credit re-anchoring rather than a generic locality or stability prior.

Temporal credit assignment is central to both biological and artificial intelligence, yet its interaction with non-linear function approximation is poorly understood. We identify a systematic failure mode in deep reinforcement learning (RL) termed Trace-Mediated Peak Bias (TMPB). At intermediate eligibility trace depths, agents irrationally prefer trajectories with high-magnitude reward peaks'' over alternatives with higher cumulative returns. This provides a mechanistic account of the Peak-End Rule: a human memory bias where experiences are judged by their most intense moments rather than integrated utility. We show that TMPB emerges because traces amplify distal Temporal Difference errors into gradient shocks'' that fixed-step-size Stochastic Gradient Descent cannot normalize, leading to global overestimation. Conversely, adaptive optimizers mitigate this pathology via second-moment normalization. Our results suggest that human-like saliency distortions may emerge naturally from the mathematical constraints of credit assignment in distributed systems, and that adaptive optimization is a theoretical necessity for rational value estimation.

Post-training via Group Relative Policy Optimization (GRPO) has emerged as a powerful paradigm for aligning flow-based generative models with human preferences. However, the iterative denoising nature of flow models incurs substantial costs when generating group rollouts for policy-gradient updates, compelling existing methods to train with extremely few denoising steps. This temporal sparsity severely restricts preference optimization: reward feedback can only reach a handful of stages per trajectory, leaving the vast majority of intermediate denoising steps without direct supervision and thus compromising alignment granularity. To address this, we propose Pave-GRPO, which reformulates the GRPO objective through Principled average velocity decomposition. Rather than generating expensive high-step rollouts, we maintain efficient few-step group sampling but decompose each coarse transition into an equivalent ensemble of finer sub-trajectories spanning multiple intermediate timesteps. This propagates reward feedback to a denser set of temporal stages for more comprehensive preference alignment without additional generation cost. This design offers two benefits: (i) zero-cost horizon expansion: through the direct reuse of piece-wise group samples and their associated rewards, Pave-GRPO significantly broadens the effective optimization scope under fixed sampling budgets; and (ii) comprehensive temporal supervision: by equivalently decomposing an instantaneous velocity target into a multi-timestep ensemble, it distributes reward signals across more intermediate stages of the denoising process, enabling finer-grained and more thorough preference optimization. Extensive experiments validate that Pave-GRPO effectively advances preference alignment across different reward settings, offering comprehensive performance enhancement.

The temporal lag between actions and their long-term consequences makes credit assignment a challenge when learning goal-directed behaviors from data. Generative world models capture the distribution of future states an agent may visit, indicating that they have captured temporal information. How can that temporal information be extracted to perform credit assignment? In this paper, we formalize how the temporal information stored in world models encodes the underlying geometry of the world. Leveraging optimal transport, we extract this geometry from a learned model of the occupancy measure into a reward function that captures goal-reaching information. Our resulting method, Occupancy Reward Shaping, largely mitigates the problem of credit assignment in sparse reward settings. ORS provably does not alter the optimal policy, yet empirically improves performance by 2.2x across 13 diverse long-horizon locomotion and manipulation tasks. Moreover, we demonstrate the effectiveness of ORS in the real world for controlling nuclear fusion on 3 Tokamak control tasks. Code: https://github.com/aravindvenu7/occupancy_reward_shaping; Website: https://aravindvenu7.github.io/website/ors/

Deploying GRPO on Flow Matching models has proven effective for text-to-image generation. However, existing paradigms typically propagate an outcome-based reward to all preceding denoising steps without distinguishing the local effect of each step. Moreover, current group-wise ranking mainly compares trajectories at matched timesteps and ignores within-trajectory dependencies, where certain early denoising actions can affect later states via delayed, implicit interactions. We propose TurningPoint-GRPO (TP-GRPO), a GRPO framework that alleviates step-wise reward sparsity and explicitly models long-term effects within the denoising trajectory. TP-GRPO makes two key innovations: (i) it replaces outcome-based rewards with step-level incremental rewards, providing a dense, step-aware learning signal that better isolates each denoising action's "pure" effect, and (ii) it identifies turning points-steps that flip the local reward trend and make subsequent reward evolution consistent with the overall trajectory trend-and assigns these actions an aggregated long-term reward to capture their delayed impact. Turning points are detected solely via sign changes in incremental rewards, making TP-GRPO efficient and hyperparameter-free. Extensive experiments also demonstrate that TP-GRPO exploits reward signals more effectively and consistently improves generation. Demo code is available at https://github.com/YunzeTong/TurningPoint-GRPO.

Working memory requires the brain to maintain information from the recent past to guide ongoing behavior. Neurons can contribute to this capacity by slowly integrating their inputs over time, creating persistent activity that outlasts the original stimulus. However, when these slowly integrating neurons are organized hierarchically, they introduce cumulative delays that create a fundamental challenge for learning: teaching signals that indicate whether behavior was correct or incorrect arrive out-of-sync with the neural activity they are meant to instruct. Here, we demonstrate that neurons enhanced with an adaptive current can compensate for these delays by responding to external stimuli prospectively -- effectively predicting future inputs to synchronize with them. First, we show that such prospective neurons enable teaching signal synchronization across a range of learning algorithms that propagate error signals through hierarchical networks. Second, we demonstrate that this successfully guides learning in slowly integrating neurons, enabling the formation and retrieval of memories over extended timescales. We support our findings with a mathematical analysis of the prospective coding mechanism and learning experiments on motor control tasks. Together, our results reveal how neural adaptation could solve a critical timing problem and enable efficient learning in dynamic environments.

The brain uses discrete spikes for dynamic computation, yet, how neural microcircuits (NMCs) solve temporal credit assignment using local spike timing remains a fundamental open question. Dominant spiking neural network (SNN) approaches circumvent this by approximating backpropagation through surrogate gradients, decoupling learning from biological spike timing. Here, we reformulate temporal credit assignment as a state separation problem: extracting task-required components induced by historical perturbations directly from the current neural state. This enables an online feedback learning framework for NMCs through a gradient tunneling (GT) algorithm and the lead-lag expansion technique that derives credit assignment from local synaptic spike timing, while remaining compatible with ANN-SNN hybrid architectures. Experimentally, GT-trained NMCs excel at long-timescale evidence integration and noise-robust memory retention, and perform comparably to leading SNN online learning methods on real-world benchmarks with far fewer parameters. The proposed framework addresses the two-decade-old NMC feedback learning problem and suggests a computationally plausible explanation for the brain's learning mechanisms.