In today’s era where datasets have exploded in size, often measuring in petabytes or streaming in real-time at velocities that defy traditional storage, linear-time algorithms that require reading every single input bit are no longer merely too slow; they are often practically infeasible. Sublinear algorithms provide the only viable path forward by intelligently sampling just a vanishing fraction of the data to yield high-quality approximate results with formally guaranteed bounds. These algorithms trade a negligible amount of accuracy for the exponential speed required to turn intractable problems into instantaneous insights.

For over two decades, sublinear algorithms have driven the development of diverse and powerful specialized tools in various computational models. The past few years have seen a rapid pace of exciting results, establishing strong connections between sublinear models and established algorithmic techniques. This progress is matched by the emergence of new paradigms incorporating crucial resources and objectives, including robustness, privacy, data erasure, interactive proofs, and oracle-aided algorithms.

The workshop will stay firmly within the theoretical realm. We plan to focus on the fundamental questions that cut across modern research into sublinear algorithms, exploring the limits of computation when we are restricted to inspecting only a small fraction of the input, storing vanishingly little information, or communicating with extremely limited bandwidth. We plan to emphasize the models, proof techniques, and conceptual trade-offs that underpin sublinear computation, while highlighting how these foundations inform a range of algorithmic domains. By bringing together researchers who study these questions from different angles, we aim to clarify the theoretical principles that allow us to make strong guarantees under such constraints.

Core themes include:

• Sublinear-time and property testing algorithms that infer global structure from a handful of random samples or local queries, highlighting how probabilistic reasoning can certify approximate answers.

• Streaming and sketching techniques that compress massive data streams into succinct summaries tailored to the target queries, spanning linear sketches as well as combinatorial structures used in graph streaming.

• Massively parallel and communication-limited graph computation models where the goal is to minimize per-machine memory or rounds of interaction, and where sublinear coordination is the central obstacle.

•  Local computation algorithms that answer queries about implicit global solutions by probing only the necessary neighborhood, clarifying the relationship between locality, randomness, and correctness.

Several overarching themes interleave these areas. Sensitivity—the extent to which a single update can affect an algorithm’s output—serves as a unifying metric: lower bounds on sensitivity translate into lower bounds on the number of rounds required in distributed and communication-limited graph algorithms. Differential privacy enters the picture through this same lens: ensuring small sensitivity is precisely what enables private mechanisms, and many of the core tools of property testing and streaming were designed to keep sensitivity low. Replicability then asks how the randomness inherent in sublinear algorithms can be tamed so that repeated runs (or independent teams) converge to comparable answers. Interactive proofs of proximity provide the complementary verification viewpoint, showing how a verifier with sublinear access still certifies global correctness via short exchanges with an untrusted prover. We will examine how these notions collectively clarify what resources sublinear computation truly needs and where its fundamental limits lie.

We plan to invite leading researchers who have made central contributions in this field as well as rising junior researchers, future stars of the field. We aim to foster in-depth discussions and creative brainstorming sessions among researchers engaged in diverse topics. We hope that the meeting will fortify collaborations across various research communities around the area of sublinear algorithms and their applications.

We will structure each session to promote coherent discussion by combining a foundational tutorial-style talk with supporting technical presentations that showcase cutting-edge research. Throughout the event, we will highlight the nascent inter-community bridges emerging from participant work. The workshop's final goal is to synthesize these findings into a shared research agenda, identifying new resource trade-offs, sharpening the limits of approximability, and establishing proof frameworks robust enough to span various computational models.

Ultimately, the goal is to distill a clear and unified story about doing more with less information. By assembling experts in sublinear-time algorithms, streaming, sketching, distributed models, privacy, and interactive verification, we expect to chart the theoretical landscape, surface crosscutting techniques, and pinpoint the questions whose answers will define the next decade of resource-efficient computation.