Omer Shlomovits

Karthik Inbasekar, Guy Rom, Omer Shlomovits

Evaluating generative video models remains an open problem. Reference-based metrics such as Structural Similarity Index Measure (SSIM) and Peak Signal to Noise Ratio (PSNR) reward pixel fidelity over semantic correctness, while Frechet Video Distance (FVD) favors distributional textures over physical plausibility. Binary Visual Question Answering (VQA) based benchmarks like VBench~2.0 are prone to yes-bias and rely on low-resolution auditors that miss temporal failures. Moreover, their prompts target a single dimension at a time, multiplying the number of videos required while still not guaranteeing reliable results. WorldJen addresses these limitations directly. Binary VQA is replaced with Likert-scale questionnaires graded by a VLM that receives frames at native video resolution. Video generation costs are addressed by using adversarially curated prompts that are designed to exercise up to 16 quality dimensions simultaneously. The framework is built around two interlocking contributions. First, A blind human preference study is conducted, accumulating (2,696 pairwise annotations from 7 annotators with 100% pair coverage over 50 of the curated prompts $\times$ 6 state-of-the-art video models. A mean inter-annotator agreement of 66.9% is achieved and the study establishes a human ground-truth Bradley-Terry (BT) rating with a three-tier structure. Second, A VLM-as-a-judge evaluation engine using prompt-specific, dimension-specific Likert questionnaires (10 questions per dimension, 47,160 scored responses) judges the videos and reproduces the human-established three-tier BT rating structure independently. The VLM achieves a Spearman $\hatρ=1.000,~p=0.0014$ that is interpreted as tier agreement with the human results. Six focused ablation studies validate the robustness of the VLM evaluation framework.

Ahmet Kurt, Suat Mercan, Omer Shlomovits et al.

Bitcoin has emerged as a revolutionary payment system with its decentralized ledger concept however it has significant problems such as high transaction fees and long confirmation times. Lightning Network (LN), which was introduced much later, solves most of these problems with an innovative concept called off-chain payments. With this advancement, Bitcoin has become an attractive venue to perform micro-payments which can also be adopted in many IoT applications (e.g. toll payments). Nevertheless, it is not feasible to host LN and Bitcoin on IoT devices due to the storage, memory, and processing requirements. Therefore, in this paper, we propose an efficient and secure protocol that enables an IoT device to use LN through an untrusted gateway node. The gateway hosts LN and Bitcoin nodes and can open & close LN channels, send LN payments on behalf of the IoT device. This delegation approach is powered by a (2,2)-threshold scheme that requires the IoT device and the LN gateway to jointly perform all LN operations which in turn secures both parties' funds. Specifically, we propose to thresholdize LN's Bitcoin public and private keys as well as its commitment points. With these and several other protocol level changes, IoT device is protected against revoked state broadcast, collusion, and ransom attacks. We implemented the proposed protocol by changing LN's source code and thoroughly evaluated its performance using a Raspberry Pi. Our evaluation results show that computational and communication delays associated with the protocol are negligible. To the best of our knowledge, this is the first work that implemented threshold cryptography in LN.

Omer Shlomovits, Oded Leiba

The blockchain space is changing constantly. New chains are being implemented frequently with different use cases in mind. As more and more types of crypto assets are getting real world value there is an increasing need for blockchain interoperability. Exchange services today are still dominated by central parties which require custody of funds. This trust imposes costs and security risks as frequent breaches testify. Atomic cross-chain swaps (ACCS) allow mutual distrusting parties to securely exchange crypto assets in a peer-to-peer manner while preserving self-custody. Fundamental ACCS protocols leveraged the scripting capabilities of blockchains to conditionalize the transfer of funds between trading parties. Recent work showed that such protocols can be realized in a scriptless setting. This has many benefits to blockchains throughput, efficiency of swap protocols and also to fungibility and privacy. The proposed protocols are limited to assets transferable by either Schnorr signatures or ECDSA that are assuming the same elliptic curve parameters. In this work we present JugglingSwap, a scriptless atomic cross-chain swap protocol with a higher degree of interoperability. We weaken the assumptions about blockchains that can be included in the ACCS protocol, and only require that (1) a threshold variant exists to the underlying digital signature scheme and (2) it is based on the elliptic curve discrete logarithm problem (ECDLP). The fair exchange is achieved by a gradual release of secrets. To achieve this we use a new building block we call Juggling: a public key verifiable encryption scheme to transfer segments of secret shares between parties, which can also be of separate interest. Juggling is then tailored to a specific private key management system design with threshold signatures security.