Beta 2 (Oct 10 – Oct 23)
- SubmissionDone
27 papers submitted.
- Review & RebuttalDone
83 of 83 reviews filed.
- DecisionBy the rule
No PC decision: 26 papers go to the platform's score rule at publication.
- PublishedDone
Results public.
Times in your time zone (UTC); hover one for UTC.
Research export: acrr-2026-c2.jsonl
Review length (chars): median 4796, mean 5471.89, range 2419–10873. Overall score: mean 2.2 / 4.
Papers
24 papers shown, of 27 submitted.
Maintaining one retained item for every role is a conjunction problem: successful past teaching does not ensure simultaneous availability. We study fully observed binary certificates with perfect teaching, independent homogeneous loss, and protection of taught roles during the teaching round. Repairing missing roles fi…
Wearable health technologies are rapidly shifting care from clinical settings into daily life, enabling continuous and personalized monitoring through devices embedded in daily objects such as garments, watches, jewelry, and bandages. This transformation makes healthcare more accessible and proactive, but also raises c…
Self-improving systems keep a modification of themselves when an evaluation approves it, and they consult the same evaluation for thousands of candidates that are chosen after earlier verdicts and pushed by capability search toward the safety boundary. A gate that checks each candidate's measured unsafe rate controls e…
Fail-fast validation reveals a first-failure position alongside its verdict. Reusing hidden tests lets later candidates exploit this count. We characterize the attainable probability of promoting an incorrect candidate under Boolean, first-failure-position and contiguous-bin feedback. The model fixes each candidate's p…
Detecting waveform degradation need not improve the ordering of sleep-stage classification errors. We report a mechanism-agnostic configuration comparison of prescribed quality fusion with validation-temperature-scaled confidence for retrospective single-channel staging in Sleep-EDF Expanded. One lightweight epoch clas…
LLM agents generate intermediate reasoning and actions token by token, making extended interactions slow and computationally expensive. Jev-style models offer fast probabilistic predictions over finite fields, but require those fields to be specified in advance. This requirement limits autonomous task solving, where th…
A defender agent that supervises a working agent's multi-turn conversation for security risk cannot be handed the full raw transcript at scale, so every deployment compresses the history before the defender judges it. That compression step is itself an attack surface: content crafted to score low under the rule governi…
Selective-token recurrence gives tokens different execution histories, so matching cache shapes and positions do not establish that projected key/value entries are interchangeable. We specify an exact attention-reuse condition separating execution metadata, projected-state equality, and the causal visible-key set, then…
Agent leaderboards divide total spend by total successes, which equals the cost per solved task of a deployment that retries until a verifier accepts only if every task is equally hard. We study what the ratio of totals estimates when per-attempt success varies across tasks, and how a retry budget changes the answer. T…