retrieval

Mixedbread introduces Toast 1, a specialized search agent that matches frontier model quality while being up to 10x cheaper and 12x faster. It automates agentic search loops and achieves state-of-the-art results on benchmarks like OfficeQA Pro V2 and legal knowledge tasks.

This paper identifies post-retrieval reuse as a bottleneck for long-horizon agent memory and proposes query-conditioned reuse (QCR), a simple target-bound note format, showing improved success and token efficiency across WebArena, WorkArena, and AppWorld.

This paper introduces ReFind, an agent-controlled search interface that works over raw, unmodified chat logs using lexical indexing and iterative keyword search, showing it can rival structured memory systems like HippoRAG 2 on conversational-memory benchmarks without building any semantic structure.

This paper presents a cost-aware comparison of LLMs versus dedicated embedding models across 37 tasks, finding that the best LLM and embedding model are nearly tied on aggregate performance but LLMs are up to 1,431x more expensive and slower, leading to a recommended division of labor.

This paper compares natively multimodal embedding models (Gemini Embedding 2, Amazon Nova 2) against frontier LLMs (GPT-4.1, Claude Sonnet 4.6) for hard-negative text-to-image retrieval, finding comparable accuracy but much lower latency for embedding-based ranking.

Presents EvoGraph-Mem, a failure-aware editable graph memory framework for long-term language agents that tracks positive/negative evidence and activation states for insights, enabling memory maintenance through utility-aware retrieval and graph-level editing.

The paper compares large language models and embedding models across 37 tasks, finding that while aggregate performance is similar, embedding models are far cheaper and faster, supporting a division of labor for cost-efficiency.

This paper systematically studies context interference in multi-turn LLM-based search agents, finding that interference primarily arises from the latest retrieved documents, and introduces a distill-based context refiner to mitigate it. Incorporating context refinement into RL training pipelines significantly improves reliability and efficiency.

This paper introduces MESA, a framework that dynamically selects and fuses a query-adaptive subset of multiple structural memory views for long-horizon agents, achieving 8.5% accuracy improvement over the strongest baseline while using 41% fewer evidence tokens.

A developer recounts how an AI feature became unexpectedly expensive under real-user usage, with long queries, repeated retrieval chunks, and unbounded conversation history, and suggests techniques like chunking and deduplication to manage token spend.

Search-G1 proposes a representation-based intrinsic reward framework for search-augmented language agents, using intervention-calibrated readouts to balance retrieval necessity and evidence reliance, improving search efficiency without costly annotations.

Introduces Controlled Memory Interference (CMI), a diagnostic framework for studying how LLM agent memory evolves under different memory relationships, revealing that relationship-specific interference suppresses update plasticity and that interference-aware training improves valid update distinction.

DistilVDR is a compact 524M visual document retriever distilled from an 8B teacher via cosine alignment, achieving near-teacher accuracy on ViDoRe with 15.6x smaller indexes and faster indexing.

The authors propose Synthetic Query Probing, a simple yet effective approach to compare different embedding models by comparing similarity match scores across models instead of raw embedding spaces. The paper illustrates non-linear relationships between models like Ada and Titan, offering practical guidance for swapping embedding models and setting retrieval thresholds.

This paper introduces SkillAligner, a training-free framework that treats retrieved skills as adaptable drafts, jointly adapting them to task requirements, execution environments, and other skills to mitigate skill-execution misfit and improve agent performance.

KReF introduces a training-free retrieval framework for long-term time-series forecasting that constructs empirical predictive distributions from similar historical lookback-future pairs, achieving strong CRPS performance across multiple benchmarks.

CertBind introduces a multiscale theory for certifiable composition of frozen multimodal encoder connectors, enabling certified retrieval decisions with fallback and abstain mechanisms. Evaluations on a shared route show it can recover native CLIP retrieval performance while preserving no-harm on passing branches.

MolBioKG is a two-layer system that grounds unseen molecules in biomedical knowledge graphs via multi-resolution structural anchoring, enabling out-of-graph link recovery and multi-hop reasoning from SMILES strings. It significantly improves Hits@10 and out-of-graph target recall over baselines while maintaining traceable evidence.

A computer science paper formalizes filesystem-based memory as long-term storage for autonomous AI agents, showing that organizing memory as hierarchical markdown files can roughly halve retrieval costs compared to vector stores.

This paper introduces Task-Conditional Flow Matching (TCFM), a framework for adapting multilingual text embedding models that selectively uses flow matching for translation tasks and other objectives for retrieval/classification. It achieves state-of-the-art results on the Indic Massive Text Embedding Benchmark.