arXiv subjects

Pedram Razavi

Publications and source records attributed to Pedram Razavi.

5 recordsLinked to original sources

arXiv2026

$\tau$-Knowledge: Evaluating Conversational Agents over Unstructured Knowledge

Conversational agents are increasingly deployed in knowledge-intensive settings, where correct behavior depends on retrieving and applying domain-specific knowledge from large, proprietary, and unstructured corpora during live interactions with users. Yet most existing benchmarks evaluate retrieval or tool use independently of each other, creating a gap in realistic, fully agentic evaluation over unstructured data in long-horizon interactions. We introduce $\tau$-Knowledge, an extension of $\tau$-Bench for evaluating agents in environments where success depends on coordinating external, natural-language knowledge with tool outputs to produce verifiable, policy-compliant state changes. Our new domain, $\tau$-Banking, models realistic fintech customer support workflows in which agents must navigate roughly 700 interconnected knowledge documents while executing tool-mediated account updates. Across embedding-based retrieval and terminal-based search, even frontier models with high reasoning budgets achieve only $\sim$25.5% pass^1, with reliability degrading sharply over repeated trials. Agents struggle to retrieve the correct documents from densely interlinked knowledge bases and to reason accurately over complex internal policies. Overall, $\tau$-Knowledge provides a realistic testbed for developing agents that integrate unstructured knowledge in human-facing deployments.

cs.AI↗

arXiv2024

$τ$-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains

Existing benchmarks do not test language agents on their interaction with human users or ability to follow domain-specific rules, both of which are vital for deploying them in real world applications. We propose $τ$-bench, a benchmark emulating dynamic conversations between a user (simulated by language models) and a language agent provided with domain-specific API tools and policy guidelines. We employ an efficient and faithful evaluation process that compares the database state at the end of a conversation with the annotated goal state. We also propose a new metric (pass^k) to evaluate the reliability of agent behavior over multiple trials. Our experiments show that even state-of-the-art function calling agents (like gpt-4o) succeed on <50% of the tasks, and are quite inconsistent (pass^8 <25% in retail). Our findings point to the need for methods that can improve the ability of agents to act consistently and follow rules reliably.

cs.AI↗

arXiv2018

Effect of strain and diameter on electronic and charge transport properties of indium arsenide nanowires

The impact of uni-axial compressive and tensile strain and diameter on the electronic band structure of indium arsenide (InAs) nanowires (NWs) is investigated using first principles calculations. Effective masses and band gaps are extracted from the electronic structure for relaxed and strained nanowires. Material properties are extracted and applied to determine charge transport through the NWs described within the effective mass approximation and by applying the non-equilibrium Green's function method. The transport calculations self-consistently solve the Schrodinger equation with open boundary conditions and Poisson's equation for the electrostatics. The device structure corresponds to a metal oxide semiconductor field effect transistor (MOSFET) with an InAs NW channel in a gate-all-around geometry. The channel cross sections are for highly scaled devices within a range of 3x3 nm2 to 1x1 nm2. Strain effects on the band structures and electrical performance are evaluated for different NW orientations and diameters by quantifying subthreshold swing and ON/OFF current ratio. Our results reveal for InAs NW transistors with critical dimensions of a few nanometer, the crystallographic orientation and quantum confinement effects dominate device behavior, nonetheless strain effects must be included to provide accurate predictions of transistor performance.

physics.app-ph↗

arXiv2018

Influence of surface stoichiometry and quantum confinement on the electronic structure of small diameter InxGa1-xAs nanowires

Electronic structures for InxGa1-xAs nanowires with [100], [110], and [111] orientations and critical dimensions of approximately 2 nm are treated within the framework of density functional theory. Explicit band structures are calculated and properties relevant to nanoelectronic design are extracted including band gaps, effective masses, and density of states. The properties of these III-V nanowires are compared to silicon nanowires of comparable dimensions as a reference system. In nonpolar semiconductors, quantum confinement and surface chemistry are known to play a key role in the determination of nanowire electronic structure. InxGa1-xAs nanowires have in addition effects due to alloy stoichiometry on the cation sublattice and due to the polar nature of the cleaved nanowire surfaces. The impact of these additional factors on the electronic structure for these polar semiconductor nanowires is shown to be significant and necessary for accurate treatment of electronic structure properties.

cond-mat.mes-hall↗

arXiv2011

On the Motion of Falling Leaves

This paper investigates the motion of falling leaves through modeling using papers and the corresponding data collected from more than four thousands experiments. Two series of experiments were designed in order to study the relationship between different parameters which can affect different paths of motion in leaves. In the first series of experiments, the shapes of the potential paths that falling papers can take were investigated as a whole. A new classification scheme was derived from these experiments, categorizing the motion of falling sheets of paper based on the deviation from the original point of release and the shape of the path they take on their descending journey. We believe this new classification scheme can be very useful with potential applications in various fields such as biology, meteorology, etc.; it can also build a foundation for further experiments. The second set of experiments was focused on the dynamics and shape of the motion of the falling paper itself. It was observed from these experiments that there should be two critical values for Froude number in the falling paper motion in regards to the rotational motion. We also found that the most stable and dominant rotational motions are those observed around the length axis as well as the ones along the diagonal axis.

physics.pop-ph↗