arXiv subjects

Pramod Viswanath

Publications and source records attributed to Pramod Viswanath.

At least 19 recordsLinked to original sources

arXiv2026

Cross-User Poisoning: User-Task Boundary Failures in Multi-User Collaborative Language Agents

Language agents are moving from single-user assistants to shared collaborators in workspaces, forums, and group chats. A shared agent observes interleaved messages from multiple users, maintains persistent context, and executes tool actions, creating a boundary problem absent from standard single-user agents: the agent must decide not only whether an instruction is safe, but which user or task it is allowed to govern. We identify cross-user poisoning (CUP), where an adversary injects a message into shared context that is later applied while the agent serves a benign user, causing unauthorized actions or responses outside the instruction's intended scope. We validate CUP on two deployed multi-user agents, Continua and ElizaOS, and introduce MURMUR, a framework for evaluating shared-context agents under concurrent multi-user interactions. Across Slack, Workspace, and Airline domains, CUP achieves high attack success, persists across later interactions, and remains substantially more effective than matched prompt-injection attacks. We evaluate boundary-scoping defenses and find that context summarization alone is insufficient, while task clustering and provenance prompting substantially reduce non-adaptive propagation. These results show that robust multi-user agents require explicit user/task scoping rather than generic input filtering or context compression.

cs.CR↗

arXiv2026

Representation-Aligned Auxiliary Supervision for Language Model Adaptation

Language models exhibit strong reasoning capabilities, yet adapting them to structured domains remains challenging and can yield inconsistent outcomes. We identify representation compatibility, the extent to which a model effectively processes a representation for a structured task, as a key factor in adaptation. We study this in chess, which provides a controlled testbed with precise semantics, computable optimal actions, and multiple state representations, including a symbolic encoding (FEN) and a spatial format (ASCII). We find that models often process semantically equivalent inputs substantially differently, affecting both learning and generalization. Building on this observation, we propose representation-aligned auxiliary supervision, which uses environment-derived tasks expressed in compatible representations to improve adaptation to structured domains. Across models and representations, auxiliary supervision consistently improves optimal-move prediction relative to target-only training under identical target data. Tasks that expose environment dynamics provide larger and most consistent gains than surface-level or static supervision, while remaining competitive with substantially increasing the amount of target-task data. Moreover, ASCII-trained models transfer more effectively to FEN than FEN-trained models do to ASCII, even surpassing the FEN target-only baseline on FEN evaluation. The gains also extend beyond optimal-move prediction to open-ended, factually grounded commentary generation. Overall, our results show that auxiliary supervision in model-compatible representations can enable effective adaptation in structured domains.

cs.CL↗

arXiv2026

Split the Yield, Share the Risk: Pricing, Hedging and Fixed rates in DeFi

We present the first formal treatment of \emph{yield tokenization}, a mechanism that decomposes yield-bearing assets into principal and yield components to facilitate risk transfer and price discovery in decentralized finance (DeFi). We propose a model that characterizes yield token dynamics using stochastic differential equations. We derive a no-arbitrage pricing framework for yield tokens, enabling their use in hedging future yield volatility and managing interest rate risk in decentralized lending pools. Taking DeFi lending as our focus, we show how both borrowers and lenders can use yield tokens to achieve optimal hedging outcomes and mitigate exposure to adversarial interest rate manipulation. Furthermore, we design automated market makers (AMMs) that incorporate a menu of bonding curves to aggregate liquidity from participants with heterogeneous risk preferences. This leads to an efficient and incentive-compatible mechanism for trading yield tokens and yield futures. Building on these foundations, we propose a modular \textit{fixed-rate} lending protocol that synthesizes on-chain yield token markets and lending pools, enabling robust interest rate discovery and enhancing capital efficiency. Our work provides the theoretical underpinnings for risk management and fixed-income infrastructure in DeFi, offering practical mechanisms for stable and sustainable yield markets.

econ.TH↗

arXiv2026

LOPMM: A Low-Order Parlay Market Maker

Parlays - joint contracts on the simultaneous resolution of several events - are among the most heavily traded products in betting markets, but prediction markets have struggled to offer them natively. The family of parlays over $M$ binary events is exponentially large, and the mechanisms that could price it either cannot be computed at that scale or need to pay for the same information multiple times. We start from an empirical observation that reframes the obstacle. On Kalshi order flow, most of the trades are placed on low-order parlays. For example, on Kalshi 7 days of traded parlay data, $87.03\%$ of the trades have at most 13 legs and $95.13\%$ have at most 20, with trade counts bounded by geometrical decay in increasing leg numbers. Given this concentration, and the structural correlation between parlays that share legs, current designs waste most of either their computation or their subsidy. We propose Low-Order Parlay Market Maker (LOPMM), which keeps a separate book for each combination of legs to stay tractable, then shares information from lower-order parlays into the higher-order ones containing them to reduce the market maker's loss. We formalize the observed concentration as a low-order flow condition. Empirically, LOPMM substantially reduces the cost of supporting parlay trading. On synthetic low-order flow, its losses remain close to that of running the base markets alone. On a replay of 83 Kalshi NBA games, independent parlay books lose \$174,050 in aggregate while LOPMM earns \$31 under the same liquidity parameter. Its computational cost depends on the order (m) of the requested parlay rather than the number (M) of base events, requiring $Θ(4^m)$ work per quote and $Θ(5^m)$ per full trade. We complement these empirical results with theoretical guarantees ...

cs.GT↗

arXiv2026

MegaFlux: Skew-Resilient MoE Megakernels via Pipelined Expert Replication

Mixture-of-experts (MoE) megakernels fuse expert-parallel communication with expert computation. However, under fixed expert placement, routing skew creates GPU stragglers: overloaded GPUs determine layer latency while others sit idle. Replicating hot experts can shift work to underloaded GPUs, but dynamic replicas introduce additional work: replicas must receive expert weights to execute and, during training, their partial weight gradients must be reduced at the expert owners. We present MegaFlux, which makes expert replication a runtime decision and pipelines the communication induced by replication within persistent MoE execution. An on-device planner jointly selects replica locations and assigns tile-aligned token blocks under a per-GPU replica budget, leaving router outputs unchanged. The forward and backward megakernels realize pipelined expert replication: replicas begin computation as their required weights arrive, while backward overlaps replica-gradient reduction with ongoing expert computation. MegaFlux extends TensorRT-LLM's CuTeDSL MegaMoE forward kernel and introduces a new backward MoE megakernel. Across 147 configurations per direction on eight NVIDIA B200 GPUs, MegaFlux achieves geometric-mean speedups of $1.45\times$ for forward and $1.28\times$ for backward over the same megakernels with fixed placement, peaking at $2.14\times$ and $2.64\times$. In ablations, pipelining hides $56$--$76$% of replica-weight transfer cost in forward and $91$--$100$% of combined weight-transfer and replica-gradient-reduction cost in backward, yielding up to $13.2$% and $26.7$% additional layer-latency reductions over the same replication plans with these operations executed separately. Integrated into vLLM for DeepSeek-V4-Pro prefill, MegaFlux delivers $1.13$--$1.26\times$ median end-to-end speedups over fixed placement.

cs.DC↗

arXiv2026

OVIG: Optimistic Verification of AI Training Integrity via Gradient Signals

The rapid growth of AI has increased the demand for domain-specific models. Post-training of open-source models offers a more economical way to meet this growing demand, but the cost of accelerator infrastructure often pushes organizations to outsource the process to third-party providers. An untrusted provider may deviate from the declared training procedure to save computation or inject malicious behavior. A potential solution is to audit the training by having an independent verifier replay the training and compare the results. This faces two key challenges: benign numerical drift from floating-point computation across heterogeneous accelerators is difficult to distinguish from malicious deviations, and the checkpoints and metadata required for verification introduce substantial storage and transmission costs. Existing approaches either have security limitations or high deployment costs. We present OVIG, an optimistic verification framework that verifies training using an empirical boundary on gradient differences calibrated from honest heterogeneous replay. Gradient differences that exceed this boundary are treated as malicious deviations. OVIG further uses optimistic verification and partitions training into stride-$s$ intervals, retaining evidence only at interval endpoints to reduce storage and transmission costs. Across shortcut training attacks and adaptive targeted manipulation attacks, OVIG maintains $0\%$ ASR on language, vision, and diffusion workloads. On Qwen3, increasing the stride from $s=1$ to $s=2000$ reduces off-chain storage and evidence transmission by $1996\times$ while preserving $0\%$ ASR and only adds $14.3\%$ system computation overhead relative to unverified training.

cs.CR↗

arXiv2026

OPFL: Optimistic Verification of Federated Learning via Empirical Boundary

Federated learning enables multiple clients to collaboratively train models without sharing their private data. However, the lack of visibility into local training makes it difficult to verify whether clients follow the prescribed training procedure or submit malicious updates, such as model poisoning. A natural approach is to replay client training for verification. However, privacy-preserving replay produces numerical results that cannot be directly matched with local client execution because the two run in different environments. We present OPFL, an optimistic verification framework for privacy-preserving federated learning. To protect data privacy, OPFL performs replay inside secure multi-party computation (MPC). Although gradients computed on MPC and local GPUs are not bitwise identical, we observe that their absolute differences are stable and bounded. OPFL therefore calibrates an empirical boundary offline and uses it to distinguish benign numerical deviations from malicious manipulation. To reduce the cost of expensive MPC replay, OPFL adopts optimistic verification by post auditing only sampled training steps. Experiments on LeNet, BERT, and Qwen show that the boundary generalizes across datasets, input lengths, and GPUs, while achieving $0$\% ASR against model poisoning and PGD-based attacks. On a LeNet workload, at $p=0.01$, OPFL is approximately $98.6\times$ faster than full MPC-based FL and $625.5\times$ faster than ZK-based approach.

cs.CR↗

arXiv2026

VeRA: Renewing Reasoning Benchmarks with Executable Specifications

Reasoning benchmarks need renewal along two axes: freshness and headroom. VeRA makes both executable and auditable by turning each item into a task family: a natural-language template, an input generator, and a deterministic answer program. VeRA-E draws fresh instances within a family; VeRA-H modifies the family toward harder tasks; and VeRA-H Pro selects one judge-ranked candidate from up to five validated proposals per seed. Execution checks, seed anchoring, answer discrimination, and independent human solving validate specifications and items. Accepted programs generate further labeled instances through local computation. Across 16 models, AIME-2024 accuracy decreases from 84.46% on seeds to 70.25% on VeRA-E variants, exposing a gap between fixed-item success and fresh-instance robustness. On AIME-2024-II, the human-audited VeRA-H Pro release lowers accuracy from 84.91% to 58.57%. Across the three hardening sources, H Pro has lower mean accuracy than H. On AMO-Bench, both releases average higher accuracy than the seeds under the evaluated budget. Initial auditing accepts 75.4% of hardened candidates; targeted repair raises usable yield to 95.1%. Executable families thus support repeatable benchmark renewal, with validation improving task quality and selection shaping the delivered challenge.

cs.AI↗

arXiv2026

AhaBench: Do Agents Turn Experience into Reusable Insights? A Long-Horizon Benchmark for Continual Learning

Can language agents continually learn from experience, turning earlier interactions into reusable capabilities? AhaBench evaluates this ability through exploration after solved hidden-state puzzles, computational transfer after mathematical teaching, and sustained business operation under delayed feedback. The benchmark is agnostic to how an agent learns; the evaluated agents use fixed model weights. Curriculum profiles, teaching contrasts, and daily trajectories reveal a common challenge: using explicit guidance is more reliable than generalizing beyond it or sustaining useful behavior. Across the Puzzle panel, the advantage over matched cold targets is 36.0-53.5 points greater with trace support than at the trace-free endpoint; Qwen 3.6 Plus nevertheless retains a +12.57-point post-curriculum gain. In Euler, worked procedures yield 80.0-100.0% held-out accuracy across models, while question-plus-answer teaching yields 0.0-73.9%. Vending trajectories separate sustained profit, late recovery, and incomplete operation: Doubao Seed 2.0 Pro finishes nominal operation at +495 but averages -10 over the year. Together, these results make continual learning an operational target: experience should yield capabilities that remain effective as guidance, inputs, and business states change. We release tasks, validators, a simulator, records, and analyses for developing agents that turn useful insights into lasting abilities.

cs.LG↗

arXiv2026

Argus: Orchestrating Cross-Layer GPU Performance Measurements around Semantic Regions

GPU developers and automated optimizers need performance evidence for semantic code regions--such as neural-network operator implementations and pipeline stages--but this evidence is fragmented across profiling tools. Answering a region-level question can require manually constructing probes and program variants, isolating interfering measurements, and mapping evidence to regions and execution contexts. We present Argus, a region-centric measurement planner and runtime that automates this workflow. Clients identify regions with boundary markers and select signals and execution scopes. Argus preserves region identity across compilation, execution, and measurement variants, constructs interference-aware multi-run plans, and orchestrates transformations and profiling across backends. It joins compiler-, hardware-, and system-level evidence using region identity and dynamic execution context, producing reports that record measurement origins and attribution ambiguity. We evaluate Argus across agentic kernel optimization, persistent megakernel optimization, and cross-level PGO. Across 44 persistent-GEMM and attention configurations, Argus improves 39/44 cases and raises AlphaEvolve's geometric-mean speedup from 5.4% to 8.9%. On a persistent TinyLlama-1.1B decode megakernel, an optimization agent reaches 1.65 ms/token with Argus versus 4.92 ms/token without it, producing a kernel $2.1\times$ faster than PyTorch with CUDA Graphs. Finally, Argus-guided cross-level PGO improves compute--communication overlap, increasing throughput by 7% on average across five multi-GPU settings.

cs.DC↗

arXiv2026

Crypto x AI, AI x Crypto: A Survey

The intersection of crypto x AI is spawning papers, products, online posts, and companies. All the surrounding buzz, though, obscures what exactly has been done, what the opportunities and challenges are, and what open questions deserve attention. This survey paper asks what AI can do for blockchain-based technologies (broadly construed as "crypto") (crypto x AI), and vice versa (AI x crypto). We systematize existing work, summarize key takeaways, highlight open research questions, and offer a perspective on pervasive industry misconceptions, concluding that AI and crypto are still in the very early stages of meaningful integration.

cs.CR↗

arXiv2026

ParlayMarket: Automated Market Making for Parlay-style Joint Contracts

Prediction markets are powerful mechanisms for information aggregation, but existing designs are optimized for single-event contracts. Traders frequently express beliefs about joint outcomes - sports parlays, conditional forecasts, multi-scenario financial bets. Current platforms either prohibit such trades or rely on ad hoc mechanisms that ignore correlation structure, resulting in inefficient prices and fragmented liquidity. We introduce \textbf{\textit{ParlayMarket}}, an automated market-maker for parlay-style joint contracts. The mechanism maintains a shared pairwise exponential-family belief state, so all base and parlay prices are marginals of one coherent distribution. This compresses the $2^M$ outcome space into $O(M^2)$ sufficient statistics and allows one liquidity pool to support an exponentially large family. Our main result characterizes the resulting learning and loss dynamics. Under repeated trading, prices converge to the best pairwise approximation of the true joint distribution. The induced expected market-maker loss grows at most quadratically in the number of base events, rather than exponentially in the number of listed parlays; moreover, this quadratic dependence is worst-case optimal for dense pairwise dependence, since there are $O(M^2)$ independent correlation directions to learn. Parlay trades are essential to this guarantee: they provide direct constraints on joint outcomes and reduce steady-state error relative to learning from marginal trades alone. Experiments on synthetic correlated markets and historical Kalshi combo data confirm the predicted scaling and show that the mechanism remains effective in realistic market-making settings. Our results demonstrate that combinatorial expressiveness does not require combinatorial capital.

cs.CE↗

arXiv2026

Correct Answers from Sound Reasoning: Verifiable Process Supervision for Language Models

Training language models to produce both correct answers and sound reasoning remains an open challenge. Reinforcement learning with verifiable rewards typically optimizes only final outcomes, which can improve task accuracy at the expense of reasoning quality, producing inaccurate, incomplete, or inconsistent traces. We propose verifiable process supervision (VPS), a post-training framework that jointly optimizes prediction accuracy and reasoning quality by supervising structured intermediate claims. We first apply supervised fine-tuning to induce a structured reasoning format, enabling deterministic extraction and verification of intermediate claims for process-level rewards. To address the heterogeneous difficulty of reasoning subtasks, we introduce adaptive weighting that prioritizes components with the largest remaining errors, creating an implicit curriculum. We evaluate VPS on chess as a controlled testbed where reasoning steps can be deterministically verified against engine signals. While outcome-only RL improves move accuracy, it sharply degrades reasoning quality, increasing win-rate error by up to 112% and reducing internal consistency by up to 69%. In contrast, VPS preserves accuracy while significantly improving reasoning quality, reducing win-rate error by up to 30% and restoring consistency to near saturation. A reasoning-space analysis further shows that, without a structured prior, outcome-only RL converges to budget-dependent shortcuts rather than sound multi-step reasoning. Beyond chess, we observe the same phenomenon on math reasoning, where outcome-only RL improves accuracy while degrading step-level arithmetic and consistency, whereas VPS maintains both. These results show that VPS enables language models to reason both accurately and reliably in verifiable domains.

cs.CL↗

arXiv2026

Hallucinations on the Board: Tool-Augmented Evaluation of LLM Chess Commentary

Superhuman game engines in domains like chess have made expert-level evaluations easily accessible, yet they communicate what is true without the natural-language explanations that make such expertise educationally useful to experts and non-experts alike. Large language models could, in principle, bridge this gap, but they frequently hallucinate due to limited domain-specific knowledge, and standard reference-based or LLM-as-a-judge frameworks cannot reliably detect these errors. In this work, we present ACT-Eval, an evaluation framework that decomposes chess commentary into atomic claims and routes them to engine-supported tools and expert-annotated gold references to assess factual correctness, conceptual coverage, and move-quality judgment. We release a benchmark of 325 position--move pairs spanning pedagogical, tournament, and critical positions, including 125 positions with expert-verified gold atoms and a five-class error taxonomy. Evaluating leading proprietary and open-weight models, we find that factual hallucinations remain pervasive in chess commentary: GPT-5.4 without tools produces incorrect sub-claims 22.0% of the time, while smaller open-weight models exceed 40%. Although tool augmentation substantially improves factual correctness and move-quality assessment, coverage of expert strategic and tactical ideas remains limited across all models. Human calibration shows that ACT-Eval's factual judgments fall within the observed range of inter-human agreement, while its coverage scores correlate strongly with human assessments of strategic completeness.

cs.CL↗

arXiv2026

Context-Aware RL for Agentic and Multimodal LLMs

Large language models (LLMs) often fail when answering requires identifying a small but decisive piece of evidence within a long or complex context, such as a single line in a tool trace or a subtle detail in an image. We propose ContextRL, a context-aware reinforcement learning (RL) method that improves long-horizon reasoning and multimodal performance through an \emph{indirect} auxiliary objective. Instead of supervising only the final answer, ContextRL presents the model with a query, an answer, and two highly similar contexts, and rewards it for selecting the context that supports the query--answer pair, thereby encouraging fine-grained grounding. We construct contrastive context data in two domains: for coding agents, trajectories serve as contexts, yielding 1k pairs built via condition filtering; for multimodal reasoning, images serve as contexts, yielding 7K pairs built via generative editing and similarity search. ContextRL achieves average gains of +2.2% over standard GRPO on 5 long-horizon benchmarks, and +1.8% across 12 diverse visual question answering benchmarks. To disentangle the effect of the proposed objective from that of additional data, we compare against data-augmentation baselines that repurpose the same contrastive contexts as standard query--context--answer examples. These baselines provide little to no improvement, showing that the gains arise from the proposed context-selection objective rather than from the contrastive data alone.

cs.CL↗

arXiv2026

TAO: Tolerance-Aware Optimistic Verification for Floating-Point Neural Networks

Neural networks increasingly run on hardware outside the user's control (cloud GPUs, inference marketplaces). Yet ML-as-a-Service reveals little about what actually ran or whether returned outputs faithfully reflect the intended inputs. Users lack recourse against service downgrades (model swaps, quantization, graph rewrites, or discrepancies like altered ad embeddings). Verifying outputs is hard because floating-point(FP) execution on heterogeneous accelerators is inherently nondeterministic. Existing approaches are either impractical for real FP neural networks or reintroduce vendor trust. We present TAO: a Tolerance Aware Optimistic verification protocol that accepts outputs within principled operator-level acceptance regions rather than requiring bitwise equality. TAO combines two error models: (i) sound per-operator IEEE-754 worst-case bounds and (ii) tight empirical percentile profiles calibrated across hardware. Discrepancies trigger a Merkle-anchored, threshold-guided dispute game that recursively partitions the computation graph until one operator remains, where adjudication reduces to a lightweight theoretical-bound check or a small honest-majority vote against empirical thresholds. Unchallenged results finalize after a challenge window, without requiring trusted hardware or deterministic kernels. We implement TAO as a PyTorch-compatible runtime and a contract layer currently deployed on Ethereum Holesky testnet. The runtime instruments graphs, computes per-operator bounds, and runs unmodified vendor kernels in FP32 with negligible overhead (0.3% on Qwen3-8B). Across CNNs, Transformers and diffusion models on A100, H100, RTX6000, RTX4090, empirical thresholds are $10^2-10^3$ times tighter than theoretical bounds, and bound-aware adversarial attacks achieve 0% success. Together, TAO reconciles scalability with verifiability for real-world heterogeneous ML compute.

cs.CR↗

arXiv2026

MINDGAMES: A Live Arena for Evaluating Social and Strategic Reasoning in Multi-Agent LLMs

Large language models (LLMs) are increasingly deployed as interactive agents, yet their capacity for social and strategic reasoning over extended interaction remains poorly understood. Existing evaluations rely on static vignettes or single-game benchmarks that cannot capture the sustained, multi-faceted reasoning that real-world multi-agent settings demand. We introduce Mindgames, a multi-game arena and evaluation platform for LLM agents that operationalizes complementary reasoning demands relevant to ``theory of mind'': belief attribution under hidden information, opponent modeling through repeated strategic interaction, cooperative inference under knowledge asymmetries, and sustained deception in social deduction. Built on TextArena, Mindgames provides a unified interaction interface, TrueSkill-based rating, and full trajectory logging across four game environments. We instantiate Mindgames through a 2025 competition cycle hosted at a major AI conference, which assessed 944 submitted agents from 76 teams across four games: Colonel Blotto, Iterated Prisoner's Dilemma, Codenames, and Secret Mafia. Our analysis surfaces both agent-level and evaluation-level limitations: brittle rule adherence remains a major bottleneck, top-performing systems repeatedly rely on explicit structural scaffolding, and leaderboard validity differs sharply across environments. In particular, failure-heavy environments can reward robustness to opponent errors as much as strategic ability, with Secret Mafia exhibiting a pronounced error-survival confound in this cycle. We release a dataset of 29,571 multi-agent games with turn-level observations, actions, and rewards, together with MG-Ref, a deterministic offline tournament protocol that scores new agents against a frozen reference pool of top-ranked, low-error Stage~II submissions under the same error-attribution lens used in this analysis.

cs.AI↗

arXiv2026

Deep-OFDM: Neural Modulation for High Mobility

Orthogonal Frequency Division Multiplexing (OFDM) is the dominant waveform in modern wireless systems, but suffers performance degradation in high-mobility environments due to Doppler-induced inter-carrier interference and unreliable pilot-based channel estimation. Neural receivers have recently shown strong performance in OFDM systems by learning equalization and detection directly from the received time-frequency grid. However, when channel estimation becomes unreliable, receiver-side learning alone is insufficient to fully recover performance. In this work we introduce DeepOFDM, a learnable modulation framework that augments conventional OFDM with a lightweight convolutional neural network (CNN) modulator jointly optimized with a neural receiver. Instead of mapping symbols independently to resource elements, DeepOFDM spreads information across local time-frequency neighborhoods while remaining fully compatible with FFT-based OFDM processing. The learned modulation breaks the rotational symmetry of conventional QAM constellations, enabling the receiver to infer residual phase directly from data symbols. This structure allows reliable operation with sparse pilots and even in fully pilotless settings. Extensive simulations demonstrate improvements in block error rate and goodput under high Doppler, while over-the-air experiments confirm practical feasibility. These results highlight the potential of transmitter-receiver co-design for robust and spectrally efficient AI-native physical layer design.

cs.IT↗