arXiv subjects

Amir Amini

Publications and source records attributed to Amir Amini.

6 recordsLinked to original sources

arXiv2026

RiskReason: A Benchmark for Verifiable Financial Risk Reasoning

Robust financial-risk inference requires more than mathematical correctness. Models must accurately apply conditional rules, maintain evolving financial states, and make decisions that respect operational constraints. This work introduces RiskReason, a benchmark for verifiable financial-risk reasoning that spans diverse risk-management mechanisms. The benchmark includes Standard tasks, which assess individual financial mechanisms, and Long-horizon tasks, which combine multiple mechanisms through extended dependencies. Each task provides reference outcomes, formal specifications, reasoning annotations, and dependency representations to support systematic analysis of model performance. Eleven model configurations are evaluated using strict single-attempt accuracy, complemented by structural and qualitative failure analyses. The strongest configurations achieve 80.0--90.2\% overall accuracy, with several maintaining high performance on composed financial scenarios. However, analysis of failures reveals incorrect risk decisions despite accurate intermediate calculations, inconsistent propagation of financial states, and remediation actions that violate binding constraints. In contrast, some strict evaluation failures result from incomplete reporting rather than incorrect financial conclusions. These findings indicate that aggregate accuracy alone is insufficient to characterize financial-risk reasoning reliability and motivate evaluations that distinguish numerical correctness, state consistency, constraint satisfaction, and decision validity.

q-fin.RM↗

arXiv2026

Certifying Reward Alignment in Portfolio Reinforcement Learning

Financial reinforcement learning often translates an investment objective into an additive risk-adjusted reward. An agent can optimise that reward accurately yet implement a different financial preference. We develop a decision audit for terminal mean-variance equilibrium that separates reward misspecification from learning error. In a finite exogenous-state market, we bound the gain from changing one action using actor, reward and continuation-critic residuals. The certificate ignores state-constant critic offsets, accommodates convex action constraints, and has an independent-sample extension that does not require the equilibrium policy. A continuation-aware actor-critic method estimates the future-policy correction from samples. Across nine controlled markets and twenty seeds, it reduces policy distance by and one-step deviation gain by relative to a matched myopic-reward network. A separate PPO comparison tests the same reward intervention within a standard optimiser. Historical tests cross seven covariance estimators and 10, 30 and 49 US industry portfolios over 432 months: joint reward calibration preserves each fitted mean-variance target, but does not improve realised investment performance. Together, these results provide a way to quantify whether a learned decision implements its stated financial objective, separately from how well that objective performs in a backtest.

q-fin.PM↗

arXiv2026

Memory Reward Inflation in Self-Improving LLM Agents

Self-improving LLM agents increasingly learn from experience without updating any weights. Each episode is stored in an external memory, scored, and retrieved for similar future tasks to shape later behavior. Viewed through a reward lens, the stored score is a proxy reward for an implicit, non-parametric policy. Each retrieved episode then becomes a policy-improvement step whose reliability hinges on how that score is produced. In deployment, ground-truth labels are unavailable, so the stored reward is at best an LLM assessment. This substitution creates a failure mode, the *Echo Gap*, across the memory-based self-improving agents and model families studied. Incorrect episodes receive inflated rewards; thus, the agent preferentially reuses the very mistakes it has most confident in. Because the error compounds through memory rather than averaging out and the confirming judge's errors remain correlated with the original self-grading bias, so it cannot identify which memories are overvalued. The missing property is formalized as the *Error-Independence Assumption* (EIA), which we prove is a *necessary* condition for correcting the inflation, not merely a description of a good verifier: a usable signal must track truth *and* decorrelate its error from the memory bias, and the recoverable payoff is a closed-form function of exactly those two quantities. We further show the inflation compounds not only when retrieval ranks by the stored score but also under plain similarity retrieval which is the regime the deployed agent uses. Finally, the answer-free de-inflation algorithm LUCID delivers a consistent end-to-end gain on the BIRD text-to-SQL benchmark. It raises execution accuracy to $56.9\%$, above both a Memento-style self-graded agent ($54.0\%$, a $+2.9$-point mean gain across seeds) and a memory-less agent of identical architecture ($52.4\%$).

cs.AI↗

arXiv2024

Secure Dynamic Event-triggered Consensus Under Asynchronous Denial of Service

This article proposes a secure implementation for consensus using a dynamic event-triggered (DET) communication scheme in high-order nonlinear multi-agent systems (MAS) under asynchronous (distributed) denial of service (DoS) attacks. By introducing a linear auxiliary trajectory of the system, the DET data transmission scheme among the neighboring agents is employed to reduce the communication for each agent. The asynchronous DoS attacks can block each communication channel among the cooperative agents independently in an unknown pattern. To guarantee state consensus of auxiliary MAS under DoS, a linear matrix inequality (LMI) based optimization approach is proposed which simultaneously designs all the unknown DET communication parameters as well as the state feedback control gain. In addition to asynchronous DoS attacks over the graph topology, the destructive effects of independent DoS attacks over the communication links between actual and auxiliary states are compensated as an additional layer of resiliency for the system. The output of each agent ultimately tracks the auxiliary state of the system and this results in the output consensus.

eess.SY↗

arXiv2019

Performance Constrained Distributed Event-triggered Consensus in Multi-agent Systems

The paper proposes a distributed eventtriggered consensus approach for linear multi-agent systems with guarantees over rate of convergence, resilience to control gain uncertainties, and Pareto optimality of design parameters, namely, the event-triggering threshold (ET) and control gain. The event-triggered consensus problem is first converted to stability problem of an equivalent system. The Lyapunov stability theorem is then used to incorporate the performance constraints with the event-triggered consensus. Using an approximated linear scalarization method, the ET and the control gain are designed simultaneously by solving a convex constrained optimization problem. Followed by some preliminary steps, the optimization can be performed locally, i.e., no global information is required. The effectiveness of the proposed approach is studied through simulations for an experimental multi-agent system.

eess.SY↗

arXiv2015

H_inf Consensus of nonlinear complex multi-agent systems using dynamic output feedback controller: An LMI approach

This paper investigates a new method for consensus in a group of nonlinear complex multi-agent systems using fixed-order non-fragile dynamic output feedback controller, via an LMI approach. The proposed scheme is decentralized in the sense that each agent relies on the relative output information among the adjacent agents. The consensus based controllers are designed to minimize the effects of nonlinear terms of the agents as well as external disturbances. Converting consensus problem to stabilization of an equivalent augmented system using proper transformations, Lyapunov stability theorem is applied to obtain unknown controller parameters in order to guarantee consensus and simultaneously acquire considered control objectives. Finally, to demonstrate the effectiveness of the proposed algorithm and compare with similar earlier researches, a numerical example on a multi-agent system consisting of single link flexible manipulators is carried out.

math.OC↗