Searcharxiv⌕ Search

arXiv subjects

Ananth Grama

Publications and source records attributed to Ananth Grama.

At least 19 recordsLinked to original sources

SARL: Label-Free Reinforcement Learning by Rewarding Reasoning Topology

Reinforcement learning is critical to improving large reasoning models, but its success relies heavily on verifiable rewards (RLVR), making it hard to use in open-ended domains where correctness is ambiguous and cannot be verified. Moreover, reasoning trajectories remain largely unconstrained, and optimizing solely toward the final answer can favor early exploitation over generalization. In this work, we ask whether general reasoning ability can be improved by teaching models how to think (the structure of reasoning) rather than what to produce (the outcome of reasoning), and we extend traditional RLVR to open-ended settings. We introduce Structure-Aware Reinforcement Learning (SARL), a label-free framework that constructs per-response reasoning maps from intermediate thinking steps and rewards their reasoning topology. SARL shifts supervision from destination to path, encouraging reasoning trajectories that are both locally coherent and globally efficient. On verifiable math tasks, SARL outperforms prior label-free RL baselines and even exceeds RL methods with ground truth supervision, with average gains of +9.1% under PPO and +11.6% under GRPO across four math benchmarks, with particularly large improvements on AIME25 (+35.5% with PPO and +44.7% with GRPO). On non-verifiable open-ended tasks, SARL achieves average gains of +34.6% under PPO and +30.4% under GRPO on WildBench across five task categories, outperforming prior label-free RL methods and DPO, which relies on additional preference labels. Beyond strong performance, SARL exhibits substantially lower KL divergence and higher policy entropy, indicating more stable and exploratory training dynamics. Code and data are available at \href{https://github.com/cacayaya/SARL}{Code Link}.

cs.AI↗

Structure makes quantum circuits trainable

The scalability of variational quantum algorithms is limited by barren plateaus, where gradients vanish exponentially in system size. This worst-case result raises fundamental questions about the trainability of quantum AI models. We present a major advance towards resolving this trainability question by answering whether a given quantum circuit has a barren plateau. We do this using an analytical framework that bounds the gradient variance for arbitrary input states $ρ$ and observables $\mathcal{M}$ by decomposing the unitary representation of the associated compact Lie group. When the circuit segments on both sides of a gate form $2$-designs, we obtain the variance exactly and prove a sufficient condition for trainability: if $ρ^{\otimes 2}$ and $\mathcal{M}^{\otimes 2}$ are sufficiently concentrated on the same irreducible invariant subspace of polynomial dimension, and the differentiated gate moves this subspace by an inverse-polynomial amount, then the gradient variance is at least inverse-polynomial. We further present a sampling algorithm that estimates the gradient variance from runs of the ansatz, without constructing the representation-theoretic projectors. Our algorithm decides whether the variance exceeds a tolerance $τ$ with the number of runs polynomial in $1/τ$, including cases in which the bounds are inconclusive. This resolves the question of the trainability of the circuit before expensive training procedures.

quant-ph↗

RestoRect: Degraded Image Restoration via Latent Rectified Flow & Feature Distillation

Current approaches for restoration of degraded images face a trade-off: high-performance models are slow for practical use, while fast models produce poor results. Knowledge distillation transfers teacher knowledge to students, but existing static feature matching methods cannot capture how modern transformer architectures dynamically generate features. We propose a novel Latent Rectified Flow Feature Distillation method for restoring degraded images called \textbf{'RestoRect'}. We apply rectified flow to reformulate feature distillation as a generative process where students learn to synthesize teacher-quality features through learnable trajectories in latent space. Our framework combines Retinex decomposition with learnable anisotropic diffusion constraints, and trigonometric color space polarization. We introduce a Feature Layer Extraction loss for robust knowledge transfer between different network architectures through cross-normalized transformer feature alignment with percentile-based outlier detection. RestoRect achieves better training stability, and faster convergence and inference while preserving restoration quality, demonstrating superior results across 15 image restoration datasets, covering 4 tasks, on 10 metrics against baselines.

cs.CV↗

LLMs Can Get "Brain Rot": A Pilot Study on Twitter/X

We propose and test the LLM Brain Rot Hypothesis: continual exposure to junk web text induces lasting cognitive decline in large language models (LLMs). To unveil junk effects, we designed a novel controlled experiment on real Twitter/X corpora, by constructing junk and reverse-controlled datasets via two orthogonal operationalizations: M1 (engagement degree) and M2 (semantic quality), with matched token scale and training operations across conditions. Compared to the control group, continual pre-training of 4 LLMs on the junk dataset causes non-trivial declines (Hedges' g>0.3) on reasoning, long-context understanding, safety, and inflating "dark traits" (e.g., psychopathy, narcissism). The gradual mixtures of junk and control datasets also yield dose-response cognition decay: for example, under M1, ARC-Challenge with Chain-of-Thought drops 72.1 -> 57.2 and RULER-CWE 83.7 -> 52.3 as junk ratio rises from 0% to 100%. Error forensics reveal several key insights. First, we identify thought-skipping as the primary lesion in reasoning: models increasingly truncate or skip chains. Second, partial but incomplete healing is observed: scaling instruction tuning and clean continual pre-training improve the declined cognition, yet cannot restore baseline capability, suggesting persistent representational drift rather than format mismatch. Finally, we discover that the popularity, a non-semantic metric, of a tweet is a better indicator of the Brain Rot effect than the length in M1. Together, the results provide significant, multi-perspective evidence that social effects of data could be a causal driver of LLM capability decay in continual pre-training, thereby motivating routine "cognitive health checks" for deployed and evolving LLMs.

cs.CL↗

Information-Theoretic Foundations of Online Temperature Sampling

Repetitive, degenerate generation is a persistent failure of LLM decoding, and the standard remedies (temperature, nucleus, and top-$k$ sampling) are static heuristics. By Kac's lemma and the AEP, a stationary source takes time exponential in its entropy to repeat a block. We therefore maximize the entropy of the sampling distributions under one cross-entropy budget that keeps the output plausible under the model. Offline, the solution is temperature sampling, with the inverse temperature as the budget's Lagrange multiplier. \emph{Our contribution is the online problem}: the conditionals $p_1, p_2, \dots$ are revealed one at a time, and the sampling distribution $q_t$ must be chosen causally from $p_1, \dots, p_t$ while the budget couples the entire horizon. The optimal temperature is thus a global shadow price that must be estimated from a prefix. We propose Entropic Online Sampling (EOS), which re-solves for the temperature on each prefix. We prove it loses $O(R\log n)$ entropy against the offline optimum while overshooting the budget by $O(R\log n)$, with $R$ an instance-dependent drift parameter. A budget-tightened variant is exactly feasible at the same rate. Conversely, every exactly feasible online algorithm must lose $Ω(R)$, so EOS is optimal up to a $\log n$ factor. When the conditionals are drawn from a finite mixture, $R=\tilde{O}(\sqrt{n})$. Experiments confirm that EOS tracks the offline temperature and beats fixed-temperature sampling in entropy.

cs.LG↗

The Optimal Decoder is Tilted Temperature

Larger sample budgets benefit from higher sampling temperatures in LLMs, but what temperature tuning optimizes and why it works remain poorly understood. We seek to answer these questions: we pose decoding as pass@$k$ maximization subject to two constraints that encode critical decoding failure modes: an entropy floor preventing diversity collapse, and a cross-entropy ceiling preventing implausibility. We prove that this maximization takes the form of a single closed-form conditional sampling rule, the \emph{tilted temperature} $q^\star \propto \bigl(p e^{ηV}\bigr)^{1/T}$. Here, the tilt comes from a verifier potential $V$ that quantifies how likely a partial generation is to eventually pass. Ordinary temperature sampling, verifier-greedy decoding, and reward-tilted policies represent special cases. We further show that, under a natural rule for scaling the diversity budget with sample size $k$, the resulting optimal temperature increases with $k$, turning empirical choice into a theorem. The optimum still needs the unknown verifier potential, which we estimate using two design approaches. In the online setting, our J-EOS algorithm jointly tracks the two constraint multipliers and learns the verifier potential from binary pass/ fail feedback alone. In the offline setting, we show that when the model is calibrated, its log-probability tracks correctness up to an affine map. Under this condition, the tilt collapses into the exponent and the optimum reduces to ordinary temperature sampling. Calibrated Temperature Decoding (CTD) then fits one scalar offline and samples with zero additional inference-time overhead relative to temperature sampling. Building on these decision rules, we design a practical decoding algorithm for code generation and evaluate it alongside CTD on benchmarks including HumanEval and MBPP, showing improvements over standard and adaptive decoding baselines.

cs.LG↗

MuonIO: Principled Norm-Aware Descent for Embedding Tables and Language Model Heads

The Muon optimizer derives its update rule for hidden linear layers by solving a local linearization of the loss penalized by the spectral norm, motivated by an RMS-stability argument for dense linear layers. Standard Muon implementations, however, exclude the input (embedding table) and output (language model head) layers from this principled treatment, for which they use AdamW instead. We present MuonIO, a single Muon-style update for both of these layers. For the language model head $\mathbf{L} \in \mathbb{R}^{V \times d}$, we motivate the use of the $2\to\infty$ operator norm, due to the Lipschitz continuity of the softmax output geometry, while for the embedding table $\mathbf{E} \in \mathbb{R}^{d \times V}$, we draw on the $1 \to 2$ operator norm, based on the one-hot input geometry identified by Bernstein & Newhouse (2025). The identity $\lVert\mathbf{L}\rVert_{2\to\infty}=\lVert\mathbf{L}^\top\rVert_{1\to2}$ then puts both matrices in the same vocabulary-oriented geometry: MuonIO applies a single normalized-vector rule, which appears as column normalization for $\mathbf{E}$ and row normalization for $\mathbf{L}$. Empirical evaluations demonstrate the effectiveness of our approach, with MuonIO reducing I/O optimizer state memory by 50% and I/O update FLOPs by $\sim$46% compared to Muon for 1B LLaMA pretraining on C4, while also improving validation perplexity.

cs.LG↗

Certified Approximation for Interpretable Representer Landmarks

Representer explanations rank the training landmarks that most influence a self-supervised representation. At scale, this ranking rests on up to four stacked approximations of the empirical neural tangent kernel (eNTK). These are random output heads, a parameter sketch, landmark sampling and a coefficient fit. Existing analyses bound each approximation separately, but none certifies the top-$K$ set against their combined error. We introduce CAIRN (Certified Approximation for Interpretable Representer laNdmarks), a framework that carries this error through to the ranking. We derive the exact variance of the sketched multi-head eNTK, which matches measurement within $4\%$ where Johnson-Lindenstrauss bounds err by up to $2.5\times$. This yields a high-probability top-$K$ certificate for a fixed coefficient fit, alongside exact residual-trace certificates for discarded spectral mass. An exact product-variance identity separates kernel error from fit variability and identifies when a larger kernel budget can still sharpen a ranking. Stochastic Lanczos Quadrature (SLQ) estimates the effective dimension within $0.72\%$ and guides the landmark budget without dense eigendecomposition. We show that residual mass does not control class coverage, and residual-greedy selection cuts the worst coverage excess of $k$-means++ from $8.5\times$ to $1.55\times$ ($4\times$ on the sketched eNTK). Cross-view initializers outperform principal-component initialization in five (AUI) to all six (CSI) settings. Against the KREPES Gauss-Newton solver, CAIRN converges $2.5$ to $11.3\times$ faster, trails by at most $0.31$ points and gains up to $3.14$ points on MNIST. Together, these results make the reliability of representer explanations measurable and show where approximation budgets are best spent.

cs.LG↗

Certified Long-Horizon Code Agent Evolution via Validation-Gated Skill Optimization

Long horizon agent self-evolution without model weight updates is essential for enabling deployed agents to accumulate reusable skills and improve over time. Prior self-evolution work has focused primarily on short-horizon tasks, while repository-level software engineering remains unexplored despite being an ideal testbed for long-horizon adaptation. In this setting, agents are required to solve streams of sequential tasks, navigate complex dependencies with evolving repositories and persistently store and reuse experience. Text-based skill optimization offers an efficient, non-parametric approach for such adaptation. However, existing methods often suffer from unstable updates, performance drawdown, and agent collapse over extended deployments. In this paper, we formalize the concept of in-context self-evolution and introduce VALVE, a validated-gated framework for long-horizon skill optimization. We establish finite convergence, provide theoretical guarantees for future-task gain and drawdown, and derive the validation and evaluation holdout sizes required for a prescribed tolerance, with leading-order scaling Empirically, our pipeline, VALVE achieves stable self-improvement over evolution horizon spanning more than 1,000 SWE tasks, with average final and peak gains of $14.9$ and $16.5$ points across three frontier models (GPT-5.5, Claude-4.6 and MiniMax-M2.7). The validation gate reduces average drawdown by 75% and produces an 11x more compact skill bank than ungated evolution. We further present extensive ablations identifying the design choices most critical to long-horizon skill evolution.

cs.AI↗

Model-Aware Data Selection from In-and-Out Information Interplay

LLMs are effective representations that assimilate vast amounts of knowledge during pretraining, but post-training is necessary for models to reliably access this knowledge and "know what they know." We observe an interesting rank equilibrium between knowledge stored in the weights and the data stream passing through the model. Across all model layers, we find that the hidden states (data stream) follow a U-shaped pattern, showing substantial compression in early layers and a steep rise during the late-layer decoding phase. In contrast, the weight rank follows an inverted U-shaped pattern, with very low rank in the early and late layers and high rank in the middle. We interpret this as an in-and-out information interplay: intermediate activations do not need to carry content that the weights can supply later, so they primarily preserve what the weights cannot provide. Motivated by this observation, we propose a model-aware data selection method, CAP (Counterfactual Assimilation Profile), which can determine whether a data candidate contains information accessible to the current model by utilizing the divergence gap in early- and late-layer representations between model-generated and reference responses. Across math, code, and science domains, CAP delivers 35.4% greater average improvement over the base model than the strongest baseline under different selection budgets. With only 10% of the data pool, CAP surpasses or matches full-pool training on math and science. We further show that CAP transfers to multimodal data selection and is robust to response horizon and noise.

cs.AI↗

Inference-Time Code Selection via Symbolic Equivalence Partitioning

Sampling multiple candidate programs at inference time is an effective way to improve LLM code generation. However, its benefit depends on reliably selecting a correct solution from the generated pool. We observe that this selection problem has a distinctive semantic structure: correct solutions, despite differences in syntax, implementation, or algorithmic strategy, often converge to the same functional behavior over valid inputs. At the same time, consensus alone is not sufficient for correctness, because models can also produce correlated wrong solutions that implement the same mistaken behavior. We propose Symbolic Equivalence Partitioning (SEP), an inference-time selection framework that first uses problem-provided public examples as lightweight validity signals. SEP then uses symbolic execution to partition the remaining candidate programs into bounded functional equivalence classes and selects from the dominant equivalence class. Across HumanEval+ and LiveCodeBench, SEP consistently improves selection accuracy without auxiliary test generation, learned verifiers, or additional LLM inference. At $N=10$, SEP improves average accuracy from 0.754 to 0.826 on HumanEval+ and from 0.565 to 0.647 on LiveCodeBench, showing that symbolic functional agreement is an effective signal for inference-time code selection.

cs.LG↗

Addressing Performance Saturation for LLM RL via Precise Entropy Curve Control

Reinforcement learning (RL) has enabled complex reasoning abilities in large language models (LLMs). However, most RL algorithms suffer from performance saturation, preventing continued gains as RL training scales. This problem can be characterized by the collapse of entropy, a key diagnostic for exploration in RL. Existing attempts focus on preventing entropy collapse through regularization or clipping. However, their resulting entropy curves often exhibit instability in the long term, which hinders performance gains. In this paper, we introduce Entrocraft, a simple rejection-sampling approach that realizes user-customized entropy schedule by biasing the advantage distributions. Entrocraft requires no objective regularization and is advantage-estimator-agnostic. Theoretically, we relate per-step entropy change to the advantage distribution under minimal assumptions. This explains the behavior of existing RL and entropy-preserving methods. Entrocraft also enables a systematic study of entropy schedules, which reveals that linear annealing, which starts high and decays to a slightly lower target, performs best. Empirically, Entrocraft addresses performance saturation, significantly improving generalization, output diversity, and long-term training. It enables a 4B model to outperform an 8B baseline, sustains improvement for up to 4x longer before plateauing, and raises pass@K by 50% over the baseline.

cs.LG↗

Unified Multi-Modal Interactive & Reactive 3D Motion Generation via Rectified Flow

Generating realistic, context-aware two-person motion conditioned on diverse modalities remains a fundamental challenge for graphics, animation and embodied AI systems. Real-world applications such as VR/AR companions, social robotics and game agents require models capable of producing coordinated interpersonal behaviour while flexibly switching between interactive and reactive generation. We introduce DualFlow, the first unified and efficient framework for multi-modal two-person motion generation. DualFlow conditions 3D motion generation on diverse inputs, including text, music, and prior motion sequences. Leveraging rectified flow, it achieves deterministic straight-line sampling paths between noise and data, reducing inference time and mitigating error accumulation common in diffusion-based models. To enhance semantic grounding, DualFlow employs a novel Retrieval-Augmented Generation (RAG) module for two-person motion that retrieves motion exemplars using music features and LLM-based text decompositions of spatial relations, body movements, and rhythmic patterns. We use a contrastive rectified flow objective to further sharpen alignment with conditioning signals and add synchronisation loss to improve inter-person temporal coordination. Extensive evaluations across interactive, reactive, and multi-modal benchmarks demonstrate that DualFlow consistently improves motion quality, responsiveness, and semantic fidelity. DualFlow achieves state-of-the-art performance in two-person multi-modal motion generation, producing coherent, expressive, and rhythmically synchronized motion.

cs.CV↗

DRIFT: Learning from Abundant User Dissatisfaction in Real-World Preference Learning

Real-world large language model deployments (e.g., conversational AI systems, code generation assistants) naturally generate abundant implicit user dissatisfaction (DSAT) signals, as users iterate toward better answers through refinements, corrections, and expressed preferences, while explicit satisfaction (SAT) feedback is scarce. Existing preference learning approaches are poorly aligned with this data profile, as they rely on costly human annotations or assume plentiful positive responses. In this paper, we introduce \textbf{DRIFT} (\textbf{D}issatisfaction-\textbf{R}efined \textbf{I}terative pre\textbf{F}erence \textbf{T}raining), which anchors training on real-world DSAT signals and samples positives dynamically from the evolving policy. Empirically, DRIFT models trained on real-world \textit{WildFeedback} datasets and synthetic \textit{UltraFeedback} datasets achieve up to +6.23\% (7B) / +7.61\% (14B) on WildBench Task Score and up to +8.95\% (7B) / +12.29\% (14B) on AlpacaEval2 win rate over base models, outperforming strong baseline methods such as iterative DPO and SPIN. At larger scales, the improvements are particularly pronounced: 14B models trained with DRIFT surpass GPT-4o-mini on WildBench. Further analysis shows that DRIFT also preserves exploratory capacity, yielding more diverse high-reward solutions rather than collapsing to narrow subsets. Theoretically, we demonstrate that this design preserves preference margins and avoids the gradient degeneration. These results show that DRIFT is an effective and scalable recipe for real-world post-training that leverages the most abundant and informative signal. The code and data are available at https://github.com/cacayaya/DRIFT.git.

cs.CL↗

Fault Oblivious Eigenvalue Solver

Eigenvalue problems serve as fundamental substrates for applications in large-scale scientific simulations and machine learning, often requiring computation on massively parallel platforms. As these platforms scale to hundreds of thousands of cores, hardware failures become a significant challenge to reliability and efficiency. In this paper, we propose and analyze a novel fault-tolerant eigenvalue solver based on erasure-coded computations -- a technique that enhances resilience by augmenting the system with redundant data a priori. This transformation reformulates the original eigenvalue problem as a generalized eigenvalue problem, enabling fault-oblivious computation while preserving numerical stability and convergence properties. We formulate the augmentation scheme, establish the necessary conditions for the encoded blocks, and prove the relationship between the original and transformed problems. We implement an erasure-coded TraceMin eigensolver and demonstrate its effectiveness in extracting eigenvalues in the presence of faults. Our experimental results show that the proposed solver incurs minimal computational overhead, maintains robust convergence, and scales efficiently with the number of faults, making it a practical solution for resilient eigenvalue computations in large-scale systems.

math.NA↗

GeneFlow: Translation of Single-cell Gene Expression to Histopathological Images via Rectified Flow

Spatial transcriptomics (ST) technologies can be used to align transcriptomes with histopathological morphology, presenting exciting new opportunities for biomolecular discovery. Using ST data, we construct a novel framework, GeneFlow, to map transcriptomics onto paired cellular images. By combining an attention-based RNA encoder with a conditional UNet guided by rectified flow, we generate high-resolution images with different staining methods (e.g. H&E, DAPI) to highlight various cellular/tissue structures. Rectified flow with high-order ODE solvers creates a continuous, bijective mapping between transcriptomics and image manifolds, addressing the many-to-one relationship inherent in this problem. Our method enables the generation of realistic cellular morphology features and spatially resolved intercellular interactions from observational gene expression profiles, provides potential to incorporate genetic/chemical perturbations, and enables disease diagnosis by revealing dysregulated patterns in imaging phenotypes. Our rectified flow-based method outperforms diffusion-based baseline method in all experiments. Code can be found at https://github.com/wangmengbo/GeneFlow.

q-bio.QM↗

Cascade Reward Sampling for Efficient Decoding-Time Alignment

Aligning large language models (LLMs) with human preferences is essential for their applications. Recently, decoding-time alignment has emerged as an effective plug-and-play technique that avoids fine-tuning model parameters. This approach retains the general utility of pretrained LLMs but often suffers from significant inefficiencies during decoding, primarily due to wasted token generation and excessive reward evaluations. To address these challenges, we introduce Cascade Reward Sampling (CARDS) to resolve both efficiency bottlenecks in decoding-time alignment. Specifically, we develop a segment-level rejection sampling algorithm that minimizes redundant computations of both LLMs and reward models (RMs). Central to CARDS is an uncertainty-based segmentation mechanism, which ensures the accuracy of RMs evaluations on incomplete segments. Furthermore, we provide a detailed analysis of reward scores on segments to elucidate the improved alignment performance. Experimental results demonstrate that CARDS significantly improves decoding efficiency, alignment quality, and general utility compared to existing decoding-time alignment methods, achieving approximately a 70% reduction in decoding time and over 90% win-ties in utility and safety benchmarks.

cs.CL↗

More is Less: The Pitfalls of Multi-Model Synthetic Preference Data in DPO Safety Alignment

Aligning large language models (LLMs) with human values is an increasingly critical step in post-training. Direct Preference Optimization (DPO) has emerged as a simple, yet effective alternative to reinforcement learning from human feedback (RLHF). Synthetic preference data with its low cost and high quality enable effective alignment through single- or multi-model generated preference data. Our study reveals a striking, safety-specific phenomenon associated with DPO alignment: Although multi-model generated data enhances performance on general tasks (ARC, Hellaswag, MMLU, TruthfulQA, Winogrande) by providing diverse responses, it also tends to facilitate reward hacking during training. This can lead to a high attack success rate (ASR) when models encounter jailbreaking prompts. The issue is particularly pronounced when employing stronger models like GPT-4o or larger models in the same family to generate chosen responses paired with target model self-generated rejected responses, resulting in dramatically poorer safety outcomes. Furthermore, with respect to safety, using solely self-generated responses (single-model generation) for both chosen and rejected pairs significantly outperforms configurations that incorporate responses from stronger models, whether used directly as chosen data or as part of a multi-model response pool. We demonstrate that multi-model preference data exhibits high linear separability between chosen and rejected responses, which allows models to exploit superficial cues rather than internalizing robust safety constraints. Our experiments, conducted on models from the Llama, Mistral, and Qwen families, consistently validate these findings.

cs.AI↗