RealICU is a new benchmark using physician hindsight labels on MIMIC-IV ICU data that exposes LLM failures in long-horizon clinical assessment, acute problem detection, action recommendation, and red-flag identification.
hub Canonical reference
Agentfold: Long-horizon web agents with proactive context management
Canonical reference. 100% of citing Pith papers cite this work as background.
28
Pith papers citing it
Background
100% of classified citations
hub tools
citation-role summary
background 5
citation-polarity summary
years
2026 28roles
background 5polarities
background 5representative citing papers
Self-programmed execution lets language models act as agents by writing and executing their own orchestrator programs in a self-modifying Lisp called Spell.
VitaBench 2.0 introduces a benchmark for long-term personalized and proactive agent behavior, with results indicating substantial gaps in current frontier LLMs.
A modular 8B agent with episodic visual memory and RL-trained retrieval reaches 91.4% cross-turn image recall over 20 turns, outperforming 32B all-context baselines with ~1.8× lower latency.
Proposes a structured concept-centric memory system for embodied agents that connects object, scene, transition, and skill memories to support coarse-to-fine retrieval and improve task performance over baselines.
MemGUI-Agent uses Context-as-Action (ConAct) for proactive context management in long-horizon GUI tasks, trained on the MemGUI-3K dataset to achieve top 8B-model results on MemGUI-Bench and MobileWorld.
HORMA builds a hierarchical memory structure from agent experiences and trains a lightweight RL navigator to retrieve minimal sufficient context, yielding better task performance with at most 22.17% of baseline token usage on ALFWorld, LoCoMo, and LongMemEval.
TreeSeeker is an inference-time tree-structured branch-and-return framework that applies textual UCB signals for controlled trial-and-error in deep search and outperforms baselines on XBench-DeepSearch and BrowseComp.
AgentFugue introduces a plug-in shared reasoning hub trained with SFT and RL that enables peer agents to share intermediate reasoning, yielding gains on long-horizon tasks over strong baselines.
SAM is a standalone memory framework for long-horizon LLM agents that creates state-adaptive cues from interactions, preserves raw trajectories for intent-driven recall, and optimizes the module via expert supervision and RL, outperforming baselines on BrowseComp and related benchmarks.
Argus coordinates a Navigator and multiple Searchers via an evidence graph for deep research, reporting average gains of 5.5 points with one Searcher and 12.7 points with eight parallel Searchers across eight benchmarks, reaching 86.2 on BrowseComp with 64 Searchers.
A single long-horizon agent with dependency-aware traversal and RepoMemory produces more complete, consistent, hierarchical repository documentation than open and closed baselines.
ToolCUA introduces a trajectory scaling pipeline and staged RL to optimize GUI-tool switching, reaching 46.85% accuracy on OSWorld-MCP for a 66% relative gain over baseline.
PruneTIR prunes erroneous tool-call trajectories during LLM inference via three trigger-based components to raise Pass@1 accuracy and efficiency while shortening context.
Context-ReAct enables agents to dynamically manage context via five atomic operations, and LongSeeker fine-tuned on 10k trajectories achieves 61.5% and 62.5% on BrowseComp benchmarks, outperforming prior agents.
An 8B multimodal search model trained with Agentic Seeding and V-Fold visual-history compression reaches state-of-the-art scores on six web-search VQA benchmarks.
LightThinker++ adds explicit adaptive memory management and a trajectory synthesis pipeline to LLM reasoning, cutting peak token use by ~70% while gaining accuracy in standard and long-horizon agent tasks.
HyMem introduces dual-granular memory storage with a lightweight summary module for fast responses and selective activation of a deep LLM module for complex queries, outperforming full-context baselines by 92.6% lower computational cost on LOCOMO and LongMemEval benchmarks.
SWE-MeM introduces adaptive memory management for coding agents via synthesized trajectories and Memory-aware GRPO, reporting 43.4% and 60.2% resolve rates on SWE-Bench Verified for 4B and 30B models while beating baselines on performance and token use.
SAFARI uses active investigation via tools and persistent short-term memory to attribute faults in agent trajectories that exceed LLM context windows, reporting 20% gains on Who&When and 0.58 precision at 5x context distance.
OSU-Mem shows overlapping memory helps retrieval when evidence shares tools or entities but hurts when steps are heterogeneous, with benefits on synthetic benchmarks vanishing on mixed real ones due to query mixing.
LRE is a kilobyte-scale learned relevance eviction policy for LLM agent memory that matches or exceeds full-history accuracy while cutting peak context size by up to 52% and using zero compressor calls.
This survey categorizes agentic environments for LLMs by eight attributes and domains, introduces symbolic and neural synthesis paradigms with evaluation, and outlines four agent evolution pathways plus three environment evolution paradigms.
ActiveMem proposes a heterogeneous distributed memory framework for LLM agents that separates planning from active memory management, reporting SOTA accuracy with lower overhead on BrowseComp-Plus and GAIA.
citing papers explorer
-
RealICU: Do LLM Agents Understand Long-Context ICU Data? A Benchmark Beyond Behavior Imitation
RealICU is a new benchmark using physician hindsight labels on MIMIC-IV ICU data that exposes LLM failures in long-horizon clinical assessment, acute problem detection, action recommendation, and red-flag identification.
-
Self-Programmed Execution for Language-Model Agents
Self-programmed execution lets language models act as agents by writing and executing their own orchestrator programs in a self-modifying Lisp called Spell.
-
VitaBench 2.0: Evaluating Personalized and Proactive Agents in Long-Term User Interactions
VitaBench 2.0 introduces a benchmark for long-term personalized and proactive agent behavior, with results indicating substantial gaps in current frontier LLMs.
-
Cognitive-structured Multimodal Agent for Multimodal Understanding, Generation, and Editing
A modular 8B agent with episodic visual memory and RL-trained retrieval reaches 91.4% cross-turn image recall over 20 turns, outperforming 32B all-context baselines with ~1.8× lower latency.
-
Analytic Concept-Centric Memory for Agentic Embodied Manipulation
Proposes a structured concept-centric memory system for embodied agents that connects object, scene, transition, and skill memories to support coarse-to-fine retrieval and improve task performance over baselines.
-
MemGUI-Agent: An End-to-End Long-Horizon Mobile GUI Agent with Proactive Context Management
MemGUI-Agent uses Context-as-Action (ConAct) for proactive context management in long-horizon GUI tasks, trained on the MemGUI-3K dataset to achieve top 8B-model results on MemGUI-Bench and MobileWorld.
-
Organize then Retrieve: Hierarchical Memory Navigation for Efficient Agents
HORMA builds a hierarchical memory structure from agent experiences and trains a lightweight RL navigator to retrieve minimal sufficient context, yielding better task performance with at most 22.17% of baseline token usage on ALFWorld, LoCoMo, and LongMemEval.
-
TreeSeeker: Tree-Structured Trial, Error, and Return in Deep Search
TreeSeeker is an inference-time tree-structured branch-and-return framework that applies textual UCB signals for controlled trial-and-error in deep search and outperforms baselines on XBench-DeepSearch and BrowseComp.
-
AgentFugue: Agent Scaling for Long-Horizon Tasks through Collective Reasoning
AgentFugue introduces a plug-in shared reasoning hub trained with SFT and RL that enables peer agents to share intermediate reasoning, yielding gains on long-horizon tasks over strong baselines.
-
SAM: State-Adaptive Memory for Long-Horizon Reasoning Agent
SAM is a standalone memory framework for long-horizon LLM agents that creates state-adaptive cues from interactions, preserves raw trajectories for intent-driven recall, and optimizes the module via expert supervision and RL, outperforming baselines on BrowseComp and related benchmarks.
-
Argus: Evidence Assembly for Scalable Deep Research Agents
Argus coordinates a Navigator and multiple Searchers via an evidence graph for deep research, reporting average gains of 5.5 points with one Searcher and 12.7 points with eight parallel Searchers across eight benchmarks, reaching 86.2 on BrowseComp with 64 Searchers.
-
Remember Your Trace: Memory-Guided Long-Horizon Agentic Framework for Consistent and Hierarchical Repository-Level Code Documentation
A single long-horizon agent with dependency-aware traversal and RepoMemory produces more complete, consistent, hierarchical repository documentation than open and closed baselines.
-
ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents
ToolCUA introduces a trajectory scaling pipeline and staged RL to optimize GUI-tool switching, reaching 46.85% accuracy on OSWorld-MCP for a 66% relative gain over baseline.
-
PruneTIR: Inference-Time Tool Call Pruning for Effective yet Efficient Tool-Integrated Reasoning
PruneTIR prunes erroneous tool-call trajectories during LLM inference via three trigger-based components to raise Pass@1 accuracy and efficiency while shortening context.
-
LongSeeker: Elastic Context Orchestration for Long-Horizon Search Agents
Context-ReAct enables agents to dynamically manage context via five atomic operations, and LongSeeker fine-tuned on 10k trajectories achieves 61.5% and 62.5% on BrowseComp benchmarks, outperforming prior agents.
-
POINTS-Seeker: An Open Recipe for Multimodal Search Agents with Visual Memory Management
An 8B multimodal search model trained with Agentic Seeding and V-Fold visual-history compression reaches state-of-the-art scores on six web-search VQA benchmarks.
-
LightThinker++: From Reasoning Compression to Memory Management
LightThinker++ adds explicit adaptive memory management and a trajectory synthesis pipeline to LLM reasoning, cutting peak token use by ~70% while gaining accuracy in standard and long-horizon agent tasks.
-
HyMem: Hybrid Memory Architecture with Dynamic Retrieval Scheduling
HyMem introduces dual-granular memory storage with a lightweight summary module for fast responses and selective activation of a deep LLM module for complex queries, outperforming full-context baselines by 92.6% lower computational cost on LOCOMO and LongMemEval benchmarks.
-
SWE-MeM: Learning Adaptive Memory Management for Long-Horizon Coding Agents
SWE-MeM introduces adaptive memory management for coding agents via synthesized trajectories and Memory-aware GRPO, reporting 43.4% and 60.2% resolve rates on SWE-Bench Verified for 4B and 30B models while beating baselines on performance and token use.
-
SAFARI: Scaling Long Horizon Agentic Fault Attribution via Active Investigation
SAFARI uses active investigation via tools and persistent short-term memory to attribute faults in agent trajectories that exceed LLM context windows, reporting 20% gains on Who&When and 0.58 precision at 5x context distance.
-
When Does Overlap Help? OSU-Mem and a Cell-Conditional Analysis of Trajectory Memory for LLM Agents
OSU-Mem shows overlapping memory helps retrieval when evidence shares tools or entities but hurts when steps are heterogeneous, with benefits on synthetic benchmarks vanishing on mixed real ones due to query mixing.
-
Learning What Not to Forget: Long-Horizon Agent Memory from a Few Kilobytes of Learning
LRE is a kilobyte-scale learned relevance eviction policy for LLM agent memory that matches or exceeds full-history accuracy while cutting peak context size by up to 52% and using zero compressor calls.
-
Agentic Environment Engineering for Large Language Models: A Survey of Environment Modeling, Synthesis, Evaluation, and Application
This survey categorizes agentic environments for LLMs by eight attributes and domains, introduces symbolic and neural synthesis paradigms with evaluation, and outlines four agent evolution pathways plus three environment evolution paradigms.
-
ActiveMem: Distributed Active Memory for Long-Horizon LLM Reasoning
ActiveMem proposes a heterogeneous distributed memory framework for LLM agents that separates planning from active memory management, reporting SOTA accuracy with lower overhead on BrowseComp-Plus and GAIA.
-
ZipRL: Adaptive Multi-Turn Context Compression with Hindsight Response Replay
ZipRL is a new RL-based adaptive compression method for multi-turn LLM agents that adds multi-granularity prompts and hindsight response replay to GRPO, reporting 27.9-34.7% gains on five agent tasks with maintained token efficiency.
-
Context Pruning for Coding Agents via Multi-Rubric Latent Reasoning
LaMR decomposes code context pruning into two rubrics using dedicated CRFs, a mixture-of-experts gate, and AST-derived labels to filter noise and often match or beat full-context baselines on coding benchmarks.
-
SWE-AGILE: A Software Agent Framework for Efficiently Managing Dynamic Reasoning Context
SWE-AGILE introduces a Dynamic Reasoning Context with sliding windows of detailed steps and compressed Reasoning Digests to enable efficient long-horizon reasoning in software engineering agents, claiming new benchmark results on SWE-Bench-Verified for 7B-8B models.
- Signal-Driven Observation for Long-Horizon Web Agents