LedgerAgent introduces a structured ledger to maintain task states separately in tool-calling agents. It renders these states into prompts and enforces policy constraints before tool execution, reducing policy violations and improving performance across customer-service domains.
This work shows that Lean can serve as a symbolic process oracle, providing fine-grained, verified feedback during reinforcement learning. By parsing proof attempts into tactic sequences and using Lean's elaboration to mark sound steps and first failures, the system generates dense, type-theoretic reward signals. Experiments demonstrate tactic-level supervision outperforms outcome-only methods on benchmarks like MiniF2F and ProofNet, highlighting Lean's role as both evaluator and training reward source.
A dual-agent framework converts natural-language experiment protocols into executable commands for robotic lab platforms. It uses a Parser Agent and a rule-based mapping engine to translate protocols, with a heterogeneous LLM Validation Agent ensuring accuracy and triggering self-correction. The framework successfully enables end-to-end autonomous execution of microplate-based experiments like the Bradford assay.
ScaffoldAgent introduces a utility-guided framework for dynamic outline optimization in open-ended deep research. It models outline evolution through Expansion, Contraction, and Revision operations, guided by a feedback mechanism that evaluates retrieval gain, structural coherence, and generation quality. Experiments show it improves long-form report generation and factual grounding compared to existing agents.
MACR introduces a multi-agent reasoning framework to resolve knowledge conflicts in LLM inference by jointly assessing internal and external knowledge. It uses semantic entropy to measure confidence and employs three specialized agents to induce rules, detect conflicts, and resolve inconsistencies across contexts. Empirical results show MACR outperforms state-of-the-art methods and provides interpretable conflict resolutions.
AutoPass uses runtime and compiler evidence to guide LLM-generated optimization decisions, outperforming expert heuristics and classical autotuning methods. It achieves geometric-mean speedups of 1.043x on x86-64 and 1.117x on ARM64 systems without prior training or fine-tuning.
VIMPO introduces a critic-free policy optimization method that derives a policy-implied value function from KL-regularized reinforcement learning. It enables verifiable reward incorporation without training a critic and outperforms GRPO on mathematical benchmarks, especially under noisy rewards.
A hierarchical system using a pretrained LLM to select RL skill policies outperforms flat RL in a 2v2 King of the Hill environment. It matches hand-crafted behavior tree performance in win rate and is perceived as more human-like by 60% of users, highlighting effective coordination and adaptability without manual rule design.
Benchmarks using fixed-shape checks miss real bugs in LLM-generated GPU kernels. A controlled corpus of 24 kernels, including 9 buggy variants with transcription errors, reveals that an op-schema-aware oracle detects all failures and passes all correct controls, with identical results across five GPU architectures.
A study evaluates four AI review systems across six language models, finding OpenAIReview with GPT-5.5 achieves 83.0% accuracy in matching paper quality to external signals and detects 71.6% of injected errors. Real user feedback shows positive sentiment, with a 1.44-to-1 vote ratio, though false positives and minor nitpicks remain common.
AgentFinVQA introduces a multi-agent pipeline for financial chart question answering that ensures auditability and on-premise deployability without significant accuracy loss. It outperforms baseline models by +7.68 pp using a proprietary backbone and +4.84 pp with open-weights Qwen3.6-27B-FP8, while providing a confidence signal via verifier output that improves human review routing.
A new framework defines when single-neuron interventions coherently control model behaviors without output collapse. The control window, based on alignment and norm ratios, predicts behavior triggers and collapse ceilings using forward pass data, with high accuracy on held-out neurons. On refusal, control is typed: coherent bypass occurs without actionable content, while genuine actionable reach appears only in specific cases and at later rollout stages.
AtomMem introduces a memory system that stores high-value atomic facts from long-form interactions. It uses hierarchical event structures and temporal profiles to capture coherent episodic contexts and track evolving user attributes, enabling stable and efficient memory evolution. Experiments on the LoCoMo benchmark show AtomMem achieves state-of-the-art performance in reasoning tasks.
REDACT introduces a systematically controlled multilingual benchmark for personally identifiable information detection, featuring 51 entity types, 4,127 surface-form patterns, and 25 languages. It evaluates five detectors across 1,000 records, revealing that rule-based models fail on high-stakes data while LLMs perform better, especially in high-sensitivity categories. A reference-free LLM assessment confirms sensitivity-tier assignment as the most challenging evaluation axis.
GEMS enables training-free superposition of multiple semantic directions in LLMs by addressing distributional deviation and directional interference through geometric constraints. On GSM8K, it maintains 98% accuracy with three non-mathematical directions, while unconstrained addition drops to 4%; on Wikitext-2, it increases PPL by only 2.2%.
LLM agents commonly select higher-privilege tools despite sufficient lower-privilege alternatives. This over-privileged behavior is amplified by transient tool failures and does not reliably improve with general safety alignment. A new privilege-aware post-training defense reduces unnecessary high-privilege tool use while maintaining agent capabilities.
Artificial Analysis' new agentic knowledge work evaluation, AA-Briefcase, shows GLM-5.2 surpassing GPT-5.5 in performance. The benchmark assesses real-world task execution and reasoning capabilities in knowledge work scenarios.
Discriminator-Guided RL (DRL) uses a pretrained representation space to train a discriminator that separates real data from model-generated samples. Its logit is used as a reward in KL-regularized RL, aligning model outputs with visual and semantic realism without human preferences. DRL improves FID and semantic FD across models like SiT and JiT, and enhances the Pareto frontier between preference and fidelity.
TxBench-PP is a verifiable benchmark for small-molecule preclinical pharmacology, testing AI agents' ability to derive accurate conclusions from real-world assay data. Across 16 model-harness configurations, no system reliably made correct preclinical pharmacology decisions, with the best performance at 59.3% (Claude Opus 4.8 / Pi) and 55.3% (GPT-5.5 / Pi) of endpoint attempts.
User as Engram proposes storing per-user facts as surgical, hash-keyed edits to a memory table, leaving reasoning in a shared adapter. This design achieves 5.6x higher indirect-reasoning accuracy and maintains base-level reasoning performance, with a memory footprint 33,000x smaller than per-user LoRA. The approach enables disjoint user edits that compose losslessly, outperforming retrieval pipelines beyond 100 facts.