arXiv subjects

Roger Wattenhofer

Publications and source records attributed to Roger Wattenhofer.

At least 73 records · Page 4Linked to original sources

arXiv2025

HyperCool: Reducing Encoding Cost in Overfitted Codecs with Hypernetworks

Overfitted image codecs like Cool-chic achieve strong compression by tailoring lightweight models to individual images, but their encoding is slow and computationally expensive. To accelerate encoding, Non-Overfitted (N-O) Cool-chic replaces the per-image optimization with a learned inference model, trading compression performance for encoding speed. We introduce HyperCool, a hypernetwork architecture that mitigates this trade-off. Building upon the N-O Cool-chic framework, HyperCool generates content-adaptive parameters for a Cool-chic decoder in a single forward pass, tailoring the decoder to the input image without requiring per-image fine-tuning. Our method achieves a 4.9% rate reduction over N-O Cool-chic with minimal computational overhead. Furthermore, the output of our hypernetwork provides a strong initialization for further optimization, reducing the number of steps needed to approach fully overfitted model performance. With fine-tuning, HEVC-level compression is achieved with 60.4% of the encoding cost of the fully overfitted Cool-chic. This work proposes a practical method to accelerate encoding in overfitted image codecs, improving their viability in scenarios with tight compute budgets.

eess.IV↗

arXiv2025

ACORD: An Expert-Annotated Retrieval Dataset for Legal Contract Drafting

Information retrieval, specifically contract clause retrieval, is foundational to contract drafting because lawyers rarely draft contracts from scratch; instead, they locate and revise the most relevant precedent. We introduce the Atticus Clause Retrieval Dataset (ACORD), the first retrieval benchmark for contract drafting fully annotated by experts. ACORD focuses on complex contract clauses such as Limitation of Liability, Indemnification, Change of Control, and Most Favored Nation. It includes 114 queries and over 126,000 query-clause pairs, each ranked on a scale from 1 to 5 stars. The task is to find the most relevant precedent clauses to a query. The bi-encoder retriever paired with pointwise LLMs re-rankers shows promising results. However, substantial improvements are still needed to effectively manage the complex legal work typically undertaken by lawyers. As the first retrieval benchmark for contract drafting annotated by experts, ACORD can serve as a valuable IR benchmark for the NLP community.

cs.CL↗

arXiv2025

Validity in Network-Agnostic Byzantine Agreement

Byzantine Agreement (BA) considers a setting of $n$ parties, out of which up to $t$ can exhibit byzantine (malicious) behavior. Honest parties must decide on a common value (agreement), which must belong to a set determined by the honest inputs (validity). Depending on the use case, this set can grow or shrink, leading to various possible desiderata collectively known as validity conditions. Varying the validity property requirement can affect the regime under which BA is solvable. Our work investigates how the selected validity property impacts BA solvability in the network-agnostic model, where the network can either be synchronous with up to $t_s$ byzantine parties or asynchronous with up to $t_a \leq t_s$ byzantine parties. We give necessary and sufficient conditions for a validity property to render BA solvable, both for the case with cryptographic setup and for the one without. This traces the precise boundary of solvability in the network-agnostic model for every validity property. Our proof of sufficiency provides a universal protocol, that achieves BA for a given validity property whenever the provided conditions are satisfied. We note that, for any non-trivial validity property, the condition $2 \cdot t_s + t_a < n$ is necessary for BA to be solvable, even with cryptographic setup. Specializing this claim to $t_a = 0$ gives that $t < n / 2$ is required whenever one expects a purely synchronous protocol to also work in an asynchronous network when there are no corruptions. This is especially surprising given that, for some validity properties, $t < n$ is a sufficient condition without the last stipulation.

cs.DC↗

arXiv2025

Mangrove: Fast and Parallelizable State Replication for Blockchains

Mangrove is a novel scaling approach to building blockchains with parallel smart contract support. Unlike in monolithic blockchains, where a single consensus mechanism determines a strict total order over all transactions, Mangrove uses separate consensus instances per smart contract, without a global order. To allow multiple instances to run in parallel while ensuring that no conflicting transactions are committed, we propose a mechanism called Parallel Optimistic Agreement. Additionally, for simple transactions, we leverage a lightweight Byzantine Reliable Broadcast primitive to reduce latency. Mangrove is optimized for performance under optimistic conditions, where there is no misbehavior and the network is synchronous. Under these conditions, our protocol can achieve a latency of 2 communication steps between creating and executing a transaction.

cs.DC↗

arXiv2025

Asynchronous Approximate Agreement with Quadratic Communication

We consider an asynchronous network of $n$ message-sending parties, up to $t$ of which are byzantine. We study approximate agreement, where the parties obtain approximately equal outputs in the convex hull of their inputs. In their seminal work, Abraham, Amit and Dolev [OPODIS '04] solve this problem in $\mathbb{R}$ with the optimal resilience $t < \frac{n}{3}$ with a protocol where each party reliably broadcasts a value in every iteration. This takes $Θ(n^2)$ messages per reliable broadcast, or $Θ(n^3)$ messages per iteration. In this work, we forgo reliable broadcast to achieve asynchronous approximate agreement against $t < \frac{n}{3}$ faults with a quadratic communication. In a tree with the maximum degree $Δ$ and the centroid decomposition height $h$, we achieve edge agreement in at most $6h + 1$ rounds with $\mathcal{O}(n^2)$ messages of size $\mathcal{O}(\log Δ+ \log h)$ per round. We do this by designing a 6-round multivalued 2-graded consensus protocol and using it to recursively reduce the task to edge agreement in a subtree with a smaller centroid decomposition height. Then, we achieve edge agreement in the infinite path $\mathbb{Z}$, again with the help of 2-graded consensus. Finally, we show that our edge agreement protocol enables $\varepsilon$-agreement in $\mathbb{R}$ in $6\log_2\frac{M}{\varepsilon} + \mathcal{O}(\log \log \frac{M}{\varepsilon})$ rounds with $\mathcal{O}(n^2 \log \frac{M}{\varepsilon})$ messages and $\mathcal{O}(n^2\log \frac{M}{\varepsilon}\log \log \frac{M}{\varepsilon})$ bits of communication, where $M$ is the maximum non-byzantine input magnitude.

cs.DC↗

arXiv2025

Recurrent Deep Differentiable Logic Gate Networks

While differentiable logic gates have shown promise in feedforward networks, their application to sequential modeling remains unexplored. This paper presents the first implementation of Recurrent Deep Differentiable Logic Gate Networks (RDDLGN), combining Boolean operations with recurrent architectures for sequence-to-sequence learning. Evaluated on WMT'14 English-German translation, RDDLGN achieves 5.00 BLEU and 30.9\% accuracy during training, approaching GRU performance (5.41 BLEU) and graceful degradation (4.39 BLEU) during inference. This work establishes recurrent logic-based neural computation as viable, opening research directions for FPGA acceleration in sequential modeling and other recursive network architectures.

cs.LG↗

arXiv2025

Transaction Fee Market Design for Parallel Execution

Given the low throughput of blockchains like Bitcoin and Ethereum, scalability - the ability to process an increasing number of transactions - has become a central focus of blockchain research. One promising approach is the parallelization of transaction execution across multiple threads. However, achieving efficient parallelization requires a redesign of the incentive structure within the fee market. Currently, the fee market does not differentiate between transactions that access multiple high-demand storage keys (i.e., unique identifiers for individual data entries) versus a single low-demand one, as long as they require the same computational effort. Addressing this discrepancy is crucial for enabling more effective parallel execution. In this work, we aim to bridge the gap between the current fee market and the need for parallel execution by exploring alternative fee market designs. To this end, we propose a framework consisting of two key components: a Gas Computation Mechanism (GCM), which quantifies the load a transaction places on the network in terms of parallelization and computation, measured in units of gas, and a Transaction Fee Mechanism (TFM), which assigns a price to each unit of gas. We additionally introduce a set of desirable properties for a GCM, propose several candidate mechanisms, and evaluate them against these criteria. Our analysis highlights two strong candidates: the weighted area GCM, which integrates smoothly with existing TFMs such as EIP-1559 and satisfies a broad subset of the outlined properties, and the time-proportional makespan GCM, which assigns gas costs based on the context of the entire block's schedule and, through this dependence on the overall execution outcome, captures the dynamics of parallel execution more accurately.

cs.GT↗

arXiv2025

Optimistic MEV in Ethereum Layer 2s: Why Blockspace Is Always in Demand

Layer 2 rollups are rapidly absorbing DeFi activity, securing over $40 billion and accounting for nearly half of Ethereum's DEX volume by Q1 2025, yet their MEV dynamics remain understudied. We address this gap by defining and quantifying optimistic MEV, a form of speculative, on-chain MEV whose detection and execution logic reside largely on-chain in smart contracts. As a result of their speculative nature and lack of off-chain opportunity verification, optimistic MEV transactions frequently decide not to execute any trades. In this work, we focus on cyclic arbitrage, which we find is predominantly executed as optimistic MEV on Layer 2s. Using our multi-stage identification pipeline on Arbitrum, Base, and Optimism, we show that in Q1 2025, transactions from cyclic arbitrage contracts account for over 50% of on-chain gas on Base and Optimism and 7% on Arbitrum, driven mainly by "interaction" probes (on-chain computations searching for arbitrage). This speculative probing indicates that cyclic arbitrage on Layer 2s is predominantly executed as optimistic MEV and contributes to generally keeping blocks on Base and Optimism persistently full. Despite consuming over half of on-chain gas, these optimistic MEV transactions pay less than one quarter of total gas fees. Cross-network comparison reveals divergent success rates, differing patterns of code reuse, and sensitivity to varying sequencer ordering and block production times. Finally, OLS regressions link optimistic MEV trade count to ETH volatility, retail trading activity, and DEX aggregator usage. Together, these findings show that optimistic MEV has become a major source of persistent spam-like transaction activity on Layer 2s, dominating blockspace with low-value probes and reshaping the composition of on-chain activity.

cs.CE↗

arXiv2025

MaskBeat: Loopable Drum Beat Generation

We present MaskBeat, a transformer-based approach for loopable drum pattern generation. Rather than predicting drum hits sequentially, our method uses bidirectional attention with iterative refinement, allowing instruments to be generated in parallel while maintaining musical coherence. Additionally, we introduce custom loss functions that capture drum-specific musical relationships. Our experiments show that MaskBeat generates higher quality and more musically coherent drum patterns than baseline approaches.

cs.SD↗

arXiv2025

Parametric Neural Amp Modeling with Active Learning

We introduce PANAMA, an active learning framework for the training of end-to-end parametric guitar amp models using a WaveNet-like architecture. With \model, one can create a virtual amp by recording samples that are determined by an active learning strategy to use a minimum amount of datapoints (i.e., amp knob settings). We show that gradient-based optimization algorithms can be used to determine the optimal datapoints to sample, and that the approach helps under a constrained number of samples.

cs.LG↗

arXiv2025

Universality Frontier for Asynchronous Cellular Automata

In this work, we investigate the computational aspects of asynchronous cellular automata (ACAs), a modification of cellular automata in which cells update independently, following an asynchronous schedule. We introduce flip automata networks (FAN), a simple modification of automata networks that remain robust under any asynchronous update schedule. We show that asynchronous automata can efficiently simulate their synchronous counterparts with a linear memory overhead, which improves upon the previously established quadratic bound. Additionally, we address the universality gap for (a)synchronous cellular automata -- the boundary separating universal and non-universal automata, which is still not fully understood. We tighten this boundary by proving that all one-way asynchronous automata lack universal computational power. Conversely, we establish the existence of a universal 6-state first-neighbor automaton in one dimension and a 3-state von Neumann automaton in two dimensions, which represent the smallest known universal constructions to date.

cs.FL↗

arXiv2025

Benchmarking Music Generation Models and Metrics via Human Preference Studies

Recent advancements have brought generated music closer to human-created compositions, yet evaluating these models remains challenging. While human preference is the gold standard for assessing quality, translating these subjective judgments into objective metrics, particularly for text-audio alignment and music quality, has proven difficult. In this work, we generate 6k songs using 12 state-of-the-art models and conduct a survey of 15k pairwise audio comparisons with 2.5k human participants to evaluate the correlation between human preferences and widely used metrics. To the best of our knowledge, this work is the first to rank current state-of-the-art music generation models and metrics based on human preference. To further the field of subjective metric evaluation, we provide open access to our dataset of generated music and human evaluations.

cs.LG↗

arXiv2025

Position Paper: Rethinking Privacy in RL for Sequential Decision-making in the Age of LLMs

The rise of reinforcement learning (RL) in critical real-world applications demands a fundamental rethinking of privacy in AI systems. Traditional privacy frameworks, designed to protect isolated data points, fall short for sequential decision-making systems where sensitive information emerges from temporal patterns, behavioral strategies, and collaborative dynamics. Modern RL paradigms, such as federated RL (FedRL) and RL with human feedback (RLHF) in large language models (LLMs), exacerbate these challenges by introducing complex, interactive, and context-dependent learning environments that traditional methods do not address. In this position paper, we argue for a new privacy paradigm built on four core principles: multi-scale protection, behavioral pattern protection, collaborative privacy preservation, and context-aware adaptation. These principles expose inherent tensions between privacy, utility, and interpretability that must be navigated as RL systems become more pervasive in high-stakes domains like healthcare, autonomous vehicles, and decision support systems powered by LLMs. To tackle these challenges, we call for the development of new theoretical frameworks, practical mechanisms, and rigorous evaluation methodologies that collectively enable effective privacy protection in sequential decision-making systems.

cs.LG↗

arXiv2025

Can Large Reasoning Models do Analogical Reasoning under Perceptual Uncertainty?

This work presents a first evaluation of two state-of-the-art Large Reasoning Models (LRMs), OpenAI's o3-mini and DeepSeek R1, on analogical reasoning, focusing on well-established nonverbal human IQ tests based on Raven's progressive matrices. We benchmark with the I-RAVEN dataset and its extension, I-RAVEN-X, which tests the ability to generalize to longer reasoning rules and ranges of the attribute values. To assess the influence of visual uncertainties on these symbolic analogical reasoning tests, we extend the I-RAVEN-X dataset, which otherwise assumes an oracle perception. We adopt a two-fold strategy to simulate this imperfect visual perception: 1) we introduce confounding attributes which, being sampled at random, do not contribute to the prediction of the correct answer of the puzzles, and 2) we smoothen the distributions of the input attributes' values. We observe a sharp decline in OpenAI's o3-mini task accuracy, dropping from 86.6% on the original I-RAVEN to just 17.0% -- approaching random chance -- on the more challenging I-RAVEN-X, which increases input length and range and emulates perceptual uncertainty. This drop occurred despite spending 3.4x more reasoning tokens. A similar trend is also observed for DeepSeek R1: from 80.6% to 23.2%. On the other hand, a neuro-symbolic probabilistic abductive model, ARLC, that achieves state-of-the-art performances on I-RAVEN, can robustly reason under all these out-of-distribution tests, maintaining strong accuracy with only a modest accuracy reduction from 98.6% to 88.0%. Our code is available at https://github.com/IBM/raven-large-language-models.

cs.AI↗

arXiv2025

Efficient Egocentric Action Recognition with Multimodal Data

The increasing availability of wearable XR devices opens new perspectives for Egocentric Action Recognition (EAR) systems, which can provide deeper human understanding and situation awareness. However, deploying real-time algorithms on these devices can be challenging due to the inherent trade-offs between portability, battery life, and computational resources. In this work, we systematically analyze the impact of sampling frequency across different input modalities - RGB video and 3D hand pose - on egocentric action recognition performance and CPU usage. By exploring a range of configurations, we provide a comprehensive characterization of the trade-offs between accuracy and computational efficiency. Our findings reveal that reducing the sampling rate of RGB frames, when complemented with higher-frequency 3D hand pose input, can preserve high accuracy while significantly lowering CPU demands. Notably, we observe up to a 3x reduction in CPU usage with minimal to no loss in recognition performance. This highlights the potential of multimodal input strategies as a viable approach to achieving efficient, real-time EAR on XR devices.

cs.CV↗

arXiv2025

Byzantine Stable Matching

In stable matching, one must find a matching between two sets of agents, commonly men and women, or job applicants and job positions. Each agent has a preference ordering over who they want to be matched with. Moreover a matching is said to be stable if no pair of agents prefer each other over their current matching. We consider solving stable matching in a distributed synchronous setting, where each agent is its own process. Moreover, we assume up to $t_L$ agents on one side and $t_R$ on the other side can be byzantine. After properly defining the stable matching problem in this setting, we study its solvability. When there are as many agents on each side with fully-ordered preference lists, we give necessary and sufficient conditions for stable matching to be solvable in the synchronous setting. These conditions depend on the communication model used, i.e., if parties on the same side are allowed to communicate directly, and on the presence of a cryptographic setup, i.e., digital signatures.

cs.DC↗

arXiv2025

Recommender Systems for Democracy: Toward Adversarial Robustness in Voting Advice Applications

Voting advice applications (VAAs) help millions of voters understand which political parties or candidates best align with their views. This paper explores the potential risks these applications pose to the democratic process when targeted by adversarial entities. In particular, we expose 11 manipulation strategies and measure their impact using data from Switzerland's primary VAA, Smartvote, collected during the last two national elections. We find that altering application parameters, such as the matching method, can shift a party's recommendation frequency by up to 105%. Cherry-picking questionnaire items can increase party recommendation frequency by over 261%, while subtle changes to parties' or candidates' responses can lead to a 248% increase. To address these vulnerabilities, we propose adversarial robustness properties VAAs should satisfy, introduce empirical metrics for assessing the resilience of various matching methods, and suggest possible avenues for research toward mitigating the effect of manipulation. Our framework is key to ensuring secure and reliable AI-based VAAs poised to emerge in the near future.

cs.CY↗

arXiv2025

Extending Asynchronous Byzantine Agreement with Crusader Agreement

In this work, we study multivalued byzantine agreement (BA) in an asynchronous network of $n$ parties where up to $t < \frac{n}{3}$ parties are byzantine. We present a new reduction from multivalued BA to binary BA. It allows one to achieve BA on $\ell$-bit inputs with one instance of binary BA, one instance of crusader agreement (CA) on $\ell$-bit inputs and $Θ(\ell n + n^2)$ bits of additional communication. As our reduction uses multivalued CA, we also design two new information-theoretic CA protocols for $\ell$-bit inputs. In the first one, we use almost-universal hashing to achieve statistical security with probability $1 - 2^{-λ}$ against $t < \frac{n}{3}$ faults with $Θ(\ell n + n^2(λ+ \log n))$ bits of communication. Following this, we replace the hashes with error correcting code symbols and add a preliminary step based on the synchronous multivalued BA protocol COOL [DISC '21] to obtain a second, perfectly secure CA protocol that can for any $\varepsilon > 0$ be set to tolerate $t \leq \frac{n}{3 + \varepsilon}$ faults with $\mathcal{O}\bigl(\frac{\ell n}{\min(1, \varepsilon^2)} + n^2\max\bigl(1, \log \frac{1}{\varepsilon}\bigr) \bigr)$ bits of communication. Our CA protocols allow one to extend binary BA to multivalued BA with a constant round overhead, a quadratic-in-$n$ communication overhead, and information-theoretic security.

cs.DC↗