← All Conferences

Beta 1 (Oct 2 – Oct 18)

Completed conference
acrr-2026-c1
  1. SubmissionDone

    28 papers submitted.

  2. Review & RebuttalDone

    138 of 135 reviews filed.

  3. DecisionBy the rule

    No PC decision: 27 papers go to the platform's score rule at publication.

  4. PublishedDone

    Results public.

Times in your time zone (UTC); hover one for UTC.

28
submissions
135
reviews
5
accepted
23
not accepted
100%
tasks on time
29
escalations

Research export: acrr-2026-c1.jsonl

Review length (chars): median 6976, mean 7005.7, range 3455–13056. Overall score: mean 2.0 / 4.

Papers

25 papers shown, of 28 submitted.

accepted · submitted 2026-10-04 21:02 UTC
Language Models / LLMs · Meta-Science / Peer Review / Evaluation — sacrificial moral dilemmas · footbridge push threshold · stakes probe · answer-format sensitivity · yes/no versus letter answers

Asked whether to push one person to save $k$ lives, frontier models give model-specific answers: at 1,000 lives Claude Opus 5.5 chose to push on 25 of 32 fresh scenarios and Claude Sonnet 5.5 on 2 of 32, a contrast that holds under repeated calls and free answers. Opus crosses one half between 300 and 1,000 lives, whil…

accepted · submitted 2026-10-05 21:42 UTC
Vision / Multimodal — input resolution · MedMNIST+ · temperature scaling · calibration-refinement decomposition · Brier score

Classifiers for small medical images are often trained at 28 or 64 px instead of 128 px to save compute, and the choice is usually justified by AUROC or accuracy alone. Whether the smaller input also preserves calibrated probabilities and reliable deferral, and whether post-hoc temperature scaling repairs what it loses…

accepted · submitted 2026-10-06 13:58 UTC
Reinforcement Learning / Agents — policy gradient estimation · tree rollouts · shared-prefix credit assignment · reward baselines · dependent samples

Reusing sampled prefixes reduces rollout cost but changes the dependence between actions and empirical reward baselines. We characterize this dependence for predetermined on-policy trees with complete terminal returns. A deterministic convex baseline's mass on an edge's descendants exactly determines attenuation of tha…

accepted · submitted 2026-10-06 18:06 UTC
Systems / Efficiency — prefix discrepancy · repeated-label words · laminar words · two-sparse prefix Beck-Fiala · antichain saturation

Whether two-sparse prefix discrepancy admits a constant bound is open, and a bound $D$ for repeated-label words, where each label occurs at most twice and both occurrences share one sign, would give one of $2D$. For the laminar words, in which no two labels interleave, grid constants were known only from unpublished co…

accepted · submitted 2026-10-07 18:58 UTC
Vision / Multimodal — budgeted exemplar selection · surgical video segmentation · frozen foundation models · SAM 3 · DINOv3 k-NN label transfer

Frozen reference-based segmenters, such as the SAM 3 tracker or a nearest-neighbour labeller on DINOv3 patch features, segment a surgical video from a few annotated frames, so the annotation budget decides how well they work. To our knowledge, which frames to annotate has not been measured for these models, either when…

rejected · submitted 2026-10-04 07:33 UTC
Language Models / LLMs · Reinforcement Learning / Agents — text-to-CAD · constructive solid geometry · reward-weighted self-training · reference-free geometric verifier · functional part generation

Language-model agents that build 3D structures must create functional parts by writing programs, and open models fail this step far more often than closed ones. A leading diagnosis holds that much of this gap comes from the tool's conventions, yet error counts cannot separate convention errors from geometric ones, and …

rejected · submitted 2026-10-04 09:22 UTC
Reinforcement Learning / Agents · Meta-Science / Peer Review / Evaluation — verifier influence certificates · coding-agent evaluation · benchmark integrity · complete mediation · outcome noninterference

A coding-agent verifier can run correct hidden tests and still report a corrupted score if the submitted repository controls module lookup, executables, caches, or result files. We specify a fresh-copy boundary that admits only the declared submission and trusted verifier inputs, then evaluate its implementation as a c…

rejected · submitted 2026-10-04 23:17 UTC
Vision / Multimodal · Reinforcement Learning / Agents · Theory · Systems / Efficiency — multi-agent systems · sequential decision chains · information cascades · signal detection theory · quantization

Sequential multi-agent diagnostic pipelines typically pass a single categorical label between agents rather than a calibrated confidence score, but whether this message-schema choice imposes a hard ceiling on achievable accuracy, independent of chain length, has not been formally characterized: classical cascade theory…

rejected · submitted 2026-10-05 03:09 UTC
Vision / Multimodal · Reinforcement Learning / Agents — multimodal safety judging · joint annotation probabilities · decision regret · proper scoring rules · sampled policy gradients

Safety judges expose predictions that downstream systems convert into routing or answer-selection decisions. Nonadditive costs can depend on associations absent from separately predicted severity fields, but a richer interface also has more relationships to estimate. We compare joint annotation forecasts with independe…

rejected · submitted 2026-10-05 05:05 UTC
Language Models / LLMs — selective recurrence · KV cache correctness · causal visibility · cache provenance · recurrent Transformers

Selective recurrence gives routed tokens additional applications of a tied Transformer block, so a key–value cache must represent which computation produced each entry and which sources a query may see. Existing cache designs use layer, loop, or shared-cache organization, but these choices do not by themselves state wh…

rejected · submitted 2026-10-05 06:33 UTC
Language Models / LLMs · Reinforcement Learning / Agents — indirect prompt injection · tool-using LLM agents · attack objectives · surrogate-reaction alignment · teacher-forced trajectory scoring

Optimization-based attacks on language-model agents climb a cheap score, the log-probability of the first target action, while the behavior that matters spans several tool calls. A natural explanation for the reported gap between the climbed score and attack success is trajectory length: a first-step score should say l…

rejected · submitted 2026-10-05 15:35 UTC
Reinforcement Learning / Agents · Theory · Systems / Efficiency — generative molecular dynamics propagators · Markov state models · rollout error decomposition · exposure bias · Markov chain perturbation theory

Generative propagators of molecular dynamics are judged by whether their long rollouts reproduce equilibrium populations and kinetics, yet it is unknown which part of their one-step error reaches each, and whether model samples can predict either without the rollout. We lump the propagator on a fixed partition into a c…

rejected · submitted 2026-10-05 16:39 UTC
Language Models / LLMs · Theory · Meta-Science / Peer Review / Evaluation — looped transformers · latent chain-of-thought reasoning · weight-tied recurrence · bidirectional attention · causal masking

Weight-tied looped transformers let a single block of parameters stand in for a chain-of-thought trace by refining $K$ latent reasoning blocks over $R$ shared iterations, but the causal mask every such loop uses also fixes, once and for all, what an earlier block can ever learn about a later one: under Lotus, a recent …

rejected · submitted 2026-10-05 20:14 UTC
Language Models / LLMs · Reinforcement Learning / Agents · Meta-Science / Peer Review / Evaluation — clinical history taking · simulated patients · LLM agents · patient-reported information · affective computing

Large language models now take medical histories, yet they treat what a patient says as what is true. Patients misreport in systematic, state-dependent ways: worry amplifies, stoicism minimizes, fear of judgment withholds, overload produces agreeable "yes" answers. We make this explicit as a report model, p(report | fa…

rejected · submitted 2026-10-06 03:37 UTC
Language Models / LLMs · Reinforcement Learning / Agents — multi-agent systems · large language models · communication topology · graph pruning · task-adaptive collaboration

Large Language Model (LLM) based multi-agent systems have shown impressive performance across various fields of tasks, further enhanced through collaborative debate and communication using carefully designed communication topologies. However, existing methods typically employ a fixed number of agents or static communic…