New research explores advanced memory and retrieval for AI agents
ByPulseAugur Editorial·
Summary by gemini-2.5-flash-lite
from 46 sources
Researchers are developing new methods to enhance the capabilities of AI agents, particularly in handling long contexts and complex reasoning tasks. Several papers propose novel approaches to memory management and retrieval, aiming to overcome limitations in current systems. These advancements include techniques for guided rereading, unified memory paradigms for network infrastructure, and benchmarks for multimodal agentic search, all contributing to more robust and efficient AI agents.
AI
Long-term memory is crucial for agents in specialized web environments, where success depends on recalling interface affordances, state dynamics, workflows, and recurring failure modes. However, existing memory benchmarks for agents mostly focus on user histories, short traces, o…
Recent advances in reinforcement learning from human feedback (RLHF) and preference optimization have substantially improved the usability, coherence, and safety of large language models. However, recurring behaviors such as performative certainty, hallucinated continuity, calibr…
Long-horizon language agents accumulate conversation history far faster than any fixed context window can hold, making memory management critical to both answer accuracy and serving cost. Existing approaches either expand the context window without addressing what is retrieved, p…
LLM-based conversational AI agents struggle to maintain coherent behavior over long horizons due to limited context. While RAG-based approaches are increasingly adopted to overcome this limitation by storing interactions in external memory modules and performing retrieval from th…
Long-horizon language agents must operate under limited runtime memory, yet existing memory mechanisms often organize experience around descriptive criteria such as relevance, salience, or summary quality. For an agent, however, memory is valuable not because it faithfully descri…
Does a lexical retriever suffice as large language models (LLMs) become more capable in an agentic loop? This question naturally arises when building deep research systems. We revisit it by pairing BM25 with frontier LLMs that have better reasoning and tool-use abilities. To supp…
To tackle long-context reasoning tasks without the quadratic complexity of standard attention mechanisms, approaches based on agent memory have emerged, which typically maintain a dynamically updated memory when linearly processing document chunks. To mitigate the potential loss …
As 6G evolves, the radio access network must transcend traditional automation to embrace agentic AI capable of perception, reasoning, and evolution. A fundamental cognitive gap persists in current disaggregated architectures, where interfaces force the physical layer to compress …
Existing benchmarks for multimodal agentic search evaluate multimodal search and visual browsing, but visual evidence is either confined to the input or treated as an answer endpoint rather than part of an interleaved search trajectory. We introduce \textbf{InterLV-Search}, a ben…
arXiv cs.CL
TIER_1·Junfeng Liao, Qizhou Wang, Jianing Zhu, Bo Du, Rui Yan, Xiuying Chen·
arXiv:2605.05583v1 Announce Type: cross Abstract: LLM agents that operate over long context depend on external memory to accumulate knowledge over time. However, existing methods typically store each observation as a single deterministic conclusion (e.g., inferring "API~X failed"…
arXiv:2605.06285v1 Announce Type: cross Abstract: Single-step retrieval-augmented generation (RAG) provides an efficient way to incorporate external information for simple question answering tasks but struggles with complex questions. Agentic RAG extends this paradigm by replacin…
arXiv cs.LG
TIER_1·Zeyu Yang, Qi Ma, Jason Chen, Anshumali Shrivastava·
arXiv:2605.06647v1 Announce Type: cross Abstract: Retrieval-augmented agents are increasingly the interface to large organizational knowledge bases, yet most still treat retrieval as a black box: they issue exploratory queries, inspect returned snippets, and iteratively reformula…
arXiv:2605.06132v1 Announce Type: new Abstract: In agent memory systems, the reranking model serves as the critical bridge connecting user queries with long-term memory. Most systems adopt the "retrieve-then-rerank" two-stage paradigm, but generic reranking models rely on semanti…
arXiv:2605.05538v1 Announce Type: new Abstract: We present AgenticRAG, a practical agentic harness for retrieval and analysis over enterprise knowledge bases. Standard RAG pipelines place significant burden of grounding on the search stack, constraining the language model to a fi…
arXiv cs.AI
TIER_1·Huyu Wu, Jun Liu, Xiaochi Wei, Yan Gao, Yi Wu, Yao Hu·
arXiv:2605.05702v1 Announce Type: new Abstract: Self-evolving search agents reduce reliance on human-written training questions by generating and solving their own search tasks. We build on Search Self-Play (SSP), a representative Proposer and Solver framework in which questions …
arXiv cs.AI
TIER_1·Zhuofeng Li, Haoxiang Zhang, Cong Wei, Pan Lu, Ping Nie, Yi Lu, Yuyang Bai, Shangbin Feng, Hangxiao Zhu, Ming Zhong, Yuyu Zhang, Jianwen Xie, Yejin Choi, James Zou, Jiawei Han, Wenhu Chen, Jimmy Lin, Dongfu Jiang, Yu Zhang·
arXiv:2605.05242v1 Announce Type: cross Abstract: Modern retrieval systems, whether lexical or semantic, expose a corpus through a fixed similarity interface that compresses access into a single top-k retrieval step before reasoning. This abstraction is efficient, but for agentic…
arXiv:2510.12635v3 Announce Type: replace Abstract: Long-context Large Language Models, despite their expanded capacity, require careful working memory management to mitigate attention dilution during long-horizon tasks. Yet existing approaches rely on external mechanisms that la…
arXiv:2604.20050v2 Announce Type: replace-cross Abstract: Can Large Language Models (AI agents) aggregate dispersed private information through trading and reason about the knowledge of others by observing price movements? We conduct a controlled experiment where AI agents trade …
Retrieval-augmented agents are increasingly the interface to large organizational knowledge bases, yet most still treat retrieval as a black box: they issue exploratory queries, inspect returned snippets, and iteratively reformulate until useful evidence emerges. This approach re…
Single-step retrieval-augmented generation (RAG) provides an efficient way to incorporate external information for simple question answering tasks but struggles with complex questions. Agentic RAG extends this paradigm by replacing single-step retrieval with a multi-step process,…
Single-step retrieval-augmented generation (RAG) provides an efficient way to incorporate external information for simple question answering tasks but struggles with complex questions. Agentic RAG extends this paradigm by replacing single-step retrieval with a multi-step process,…
In agent memory systems, the reranking model serves as the critical bridge connecting user queries with long-term memory. Most systems adopt the "retrieve-then-rerank" two-stage paradigm, but generic reranking models rely on semantic similarity matching and lack genuine reasoning…
arXiv:2605.04897v1 Announce Type: new Abstract: Extraction at ingestion is the wrong primitive for agent memory: content discarded before the query is known cannot be recovered at retrieval time. We propose True Memory, a six-layer architecture that shifts the center of the syste…
Long-horizon search agents must manage a rapidly growing working context as they reason, call tools, and observe information. Naively accumulating all intermediate content can overwhelm the agent, increasing costs and the risk of errors. We propose that effective context manageme…
Extraction at ingestion is the wrong primitive for agent memory: content discarded before the query is known cannot be recovered at retrieval time. We propose True Memory, a six-layer architecture that shifts the center of the system from a storage schema to a multi-stage retriev…
arXiv:2605.02491v1 Announce Type: cross Abstract: Modern searches for physics beyond the Standard Model produce rapidly expanding literature containing heterogeneous information, including textual analyses, numerical datasets, and graphical exclusion limits. Integrating these dis…
arXiv:2605.04018v1 Announce Type: new Abstract: Reasoning-intensive retrieval aims to surface evidence that supports downstream reasoning rather than merely matching topical similarity. This capability is increasingly important for agentic search systems, where retrievers must pr…
Reasoning-intensive retrieval aims to surface evidence that supports downstream reasoning rather than merely matching topical similarity. This capability is increasingly important for agentic search systems, where retrievers must provide complementary evidence across iterative se…
Modern searches for physics beyond the Standard Model produce rapidly expanding literature containing heterogeneous information, including textual analyses, numerical datasets, and graphical exclusion limits. Integrating these distributed sources remains a time-consuming and manu…
Modern searches for physics beyond the Standard Model produce rapidly expanding literature containing heterogeneous information, including textual analyses, numerical datasets, and graphical exclusion limits. Integrating these distributed sources remains a time-consuming and manu…
<p>Every AI agent you build today can hold a conversation. It can reason, use tools, and chain together complex workflows. But the moment a session ends, everything disappears. The agent forgets who you are, what you were working on, and every preference it learned during the con…
<h2> The Memory Problem in AI Agents </h2> <p>Modern LLMs are incredibly powerful, but they have a fundamental limitation: <strong>they forget everything between conversations</strong>. Every time you start a new session with an AI agent, it's like talking to someone with amnesia…
<p>I kept running into the same problem with AI coding agents.</p> <p>The agents were getting better, but every new session still felt like starting<br /> from zero.</p> <p>I would explain the repo again. Then my preferences again. Then the decisions we<br /> already made. Then w…
<p>I kept running into the same problem with AI coding agents.</p> <p>The agents were getting better, but every new session still felt like starting<br /> from zero.</p> <p>I would explain the repo again. Then my preferences again. Then the decisions we<br /> already made. Then w…
<h2>From Stateless Prompts to Persistent Intelligence</h2> <blockquote> <strong>Where this fits:</strong> This article bridges two series. It closes out the themes introduced in The Backyard Quarry — a data engineering exploration using physical objects as a teaching domain — and…
🧠 Graft provides a semantic memory system for AI agents that operates independently of large language models. The tool allows agents to store and retrieve information based on meaning rather than exact text matching. 💬 Hacker News 🔗 https:// github.com/AEndrix03/Graft # AI # Mach…
<p>In the world of Large Language Models (LLMs), we often face a frustrating paradox: LLMs are incredibly capable at "reasoning" in the moment, but they are fundamentally <strong>stateless</strong>. Every time you start a new session, the agent has total amnesia. It doesn't remem…
<p><em>Originally published on <a href="https://www.poniaktimes.com/subq-model-efficient-long-context-ai/" rel="noopener noreferrer">Poniak Times</a>. Reposted here for the developer and AI engineering community.</em></p> <p>Subquadratic’s SubQ model claims to make long-context A…
<p>If you are building agents in 2026, you have already hit the wall. Bigger models do not fix forgetfulness. Context windows can grow forever, and the agent still cannot remember what a user told it last Tuesday, that the customer's address changed three months ago, or that a re…
<blockquote> <p><em>This article was originally published on <a href="https://dingjiu1989-hue.github.io/en/ai/ai-agents-memory-patterns.html" rel="noopener noreferrer">AI Study Room</a>. For the full version with working code examples and related articles, visit the original post…
<blockquote> <p>English version: <a href="https://dev.to/tirsogarcia/building-kernel-memory-protocol-navigable-memory-for-ai-agents-315j">Building Kernel Memory Protocol: Navigable Memory for AI Agents</a></p> </blockquote> <p>El problema de muchos agentes de IA no es que les fal…
<blockquote> <p>Versión en español: <a href="https://dev.to/tirsogarcia/construyendo-kernel-memory-protocol-memoria-navegable-para-agentes-de-ia-24lc">Construyendo Kernel Memory Protocol: memoria navegable para agentes de IA</a></p> </blockquote> <p>The hard part with many AI age…
<h1> How Agentic Search Actually Works: The Research Loop Link-Fetching Agents Miss </h1> <p>Most agent tutorials show you the same pattern: take a user query, call a search API, grab the top result, stuff the text into your prompt. Done. Ship it.</p> <p>That works fine for trivi…