5arXiv cs.CL (Computation and Language)·
GRACE: Graph-Regularized Agentic Context Evolution for reliable long-horizon instruction updates
Researchers introduce GRACE, a method that maintains a deployed LLM agent's persistent system-level instructions as a typed semantic graph rather than flat text, enabling local verification of updates within typed node neighborhoods. Evaluated on a telecom agent harness derived from τ²-bench under distribution shift, GRACE improves pass³ reliability from 0.091 (Gemini 2.5 Flash zero-shot) to 0.673±0.136, surpassing a Gemini 3.1 Pro zero-shot reference of 0.242. The work identifies structural substrate and consolidation mechanisms as key requirements for reliable long-horizon agentic context evolution. The flat-text baseline finishes at 0.191, underscoring the practical gap GRACE addresses.
Related guides (3)
Related events (8)
CompactionRL trains long-horizon agents with context compaction via reinforcement learning
Researchers propose CompactionRL, a reinforcement learning strategy that jointly optimizes task execution and context summarization to enable LLM agents to operate beyond finite context windows. The method uses token-level loss normalization and cross-trajectory generalized advantage estimation to learn from compacted long-horizon trajectories. Applied to open GLM models, CompactionRL achieves 66.8% Pass@1 on SWE-bench Verified with GLM-4.5-Air (106B-A30B), a 7.0-point absolute gain, and has been incorporated into the training pipeline for GLM-5.2 (750B-A40B).
HULAT2 multi-agent LangGraph system for Spanish Easy-to-Read text simplification at MER-TRANS 2026
Researchers from HULAT2-UC3M describe their submission to the MER-TRANS 2026 shared task on multilingual Easy-to-Read translation, using a LangGraph-based multi-agent workflow combining Gemini 2.5 Flash and RigoChat-7B-v2. The best run (RUN1) achieved a SARI score of 44.05 using Event-Condition-Action routing and internal quality signals, outperforming a LoRA-adapted generate-evaluate-regenerate baseline. Results show signal-guided multi-agent routing outperforms linear regeneration, while adding lexical support did not automatically improve reference-based scores.
STRACE framework improves LLM agent optimization via structural trajectory analysis and causal extraction
Researchers introduce STRACE (Structural TRajectory Analysis and Causal Extraction), a framework for constructing high signal-to-noise optimization contexts for long-horizon LLM agents. At the batch level, STRACE mines failure patterns to filter redundant traces; within each trace, it performs causal localization over a textual dependency graph to isolate root-cause steps. On the formal verification benchmark VeruSAGE-Bench, STRACE achieves a 1.4× success-rate improvement (42.5% to 58.5%) over human-expert-designed agents, outperforming standard context-filtering baselines.
REAL: Reasoning-enhanced temporal graph framework for LLM long-term memory management
REAL is a new framework that represents LLM conversational memory as a temporal, confidence-aware directed property graph, where atomic facts carry validity intervals, confidence scores, and exploration intent labels. It addresses three limitations of prior memory systems: flat text structures, destructive overwrites of evolving facts, and passive retrieval. The system uses non-destructive temporal updates, semantic evaluator-guided hybrid beam search, and counterfactual inference to repair incomplete retrieval states. Experiments show a 22.72% average improvement over flat-text, graph-based, and existing memory baselines.
FORGE: Self-Evolving Agent Memory via Population Broadcast Without Weight Updates
FORGE (Failure-Optimized Reflective Graduation and Evolution) is a staged, population-based protocol that evolves prompt-injected natural-language memory for hierarchical ReAct agents without any gradient updates. It wraps a Reflexion-style inner loop where a reflection agent converts failed trajectories into textual heuristics or few-shot demonstrations, then propagates the best-performing instance's memory across a population between stages. Evaluated on CybORG CAGE-2 (a stochastic network-defense POMDP), FORGE improves average return by 1.7–7.7× over zero-shot and 29–72% over Reflexion across all 12 model-representation conditions tested with four LLM families. Notably, weaker models benefit disproportionately, suggesting the method may help close capability gaps rather than amplify already-strong models.
Agentic CLEAR: Automating Multi-Level Evaluation of LLM Agents
Agentic CLEAR is an automatic evaluation framework for LLM-based agentic systems that analyzes behavior at three granularity levels: system, trace, and node. Unlike existing tools that rely on static error taxonomies or focus only on observability, it dynamically generates textual insights and integrates above the observability layer with an accessible UI. Experiments across four benchmarks and seven agentic settings demonstrate strong alignment with human-annotated errors and predictive accuracy for task success rates.
LedgerAgent: Structured state ledger improves policy-adherent tool-calling agents
LedgerAgent is an inference-time method that maintains explicit task state in a separate ledger rather than leaving state reconstruction implicit in the prompt, addressing two failure modes: stale/incorrect grounding and policy-violating tool calls. The ledger is used both to render current state into the prompt and to gate environment-changing tool calls against state-dependent policy constraints. Evaluated across four customer-service domains with a mixed panel of open- and closed-weight models, LedgerAgent improves average pass^k over standard prompt-based tool-calling, with the largest gains under stricter multi-trial consistency metrics.
ContextRL: Context-aware reinforcement learning improves grounding in agentic and multimodal LLMs
Researchers introduce ContextRL, a reinforcement learning method that trains LLMs to select the context that supports a given query-answer pair from two highly similar candidates, rather than supervising only final answers. The approach constructs contrastive context pairs in two domains: coding agent trajectories (1k pairs) and multimodal image pairs (7k pairs). ContextRL achieves +2.2% average gains over standard GRPO on 5 long-horizon benchmarks and +1.8% across 12 visual QA benchmarks, with ablations showing the gains stem from the context-selection objective rather than the contrastive data alone.