The paper introduces MORSE, a method that orders contexts in a way that preserves evidence during compression. By applying a reverse query‑evidence principle to both individual contexts and compressed outputs, MORSE constructs an evidence‑first anchor and guides permutation selection to avoid information preemption. Experiments on multi‑hop QA benchmarks show that MORSE consistently outperforms static reverse ordering and random search, improving evidence retention and downstream QA performance.
arXiv:2605.28093v3 Announce Type: replace Abstract: Retrieval-augmented generation (RAG) has emerged as a promising paradigm for enhancing large language models (LLMs) on multi-hop question answering...
CacheWeaver is a lightweight prompt‑layer technique that orders evidence for Retrieval‑Augmented Generation (RAG) to improve cache reuse in serving engines like vLLM. By maintaining a prefix tree of recently served evidence sequences and greedily placing the most reusable prefix first, it reduces median time‑to‑first‑token by 20‑33 % across three vLLM configurations without harming answer quality. The greedy policy achieves 97.5 % of the gain possible with oracle ordering, showing that most reusable prefix locality can be recovered with a simple scheduling layer.
MiniVer‑V is a new benchmark for short‑video fact‑checking that contains 195 videos, each annotated with a supported, refuted, or insufficient verdict and 5,510 multimodal evidence units (visual keyframes, speech transcripts, and web‑retrieved sources). The authors propose a two‑layer verification framework that first checks claim‑video consistency using internal evidence, then determines the factual verdict with external corroboration, and employs a sufficiency‑driven greedy search to assemble only the minimal evidence needed for a confident decision. Using Claude Sonnet 4, the method achieves a Macro‑F1 of 0.510 with an average of 4.5 evidence units—about 16 % of the full set—while matching the full‑evidence baseline and improving detection of insufficient cases.
arXiv:2606. 00432v1 Announce Type: new Abstract: As retrieval-augmented generation (RAG) systems scale, it becomes increasingly challenging to ensure faithful grounding in external evidence.
arXiv:2608.23011v1 Announce Type: cross Abstract: Graph-based retrieval-augmented generation (RAG) provides a scalable paradigm for long-video understanding, but existing systems typically inherit a...
Coverage Under Contradiction: A Benchmark for Negation-Aware Retrieval introduces CoverageBench-Negate, a new benchmark that augments the existing CoverageBench with controlled contradictions. For each evidence nugget, a negated counterpart is created and its supporting documents are edited to support the negation while preserving context, resulting in about 63,000 nugget‑document pairs across six datasets. The benchmark includes rich construction metadata, LLM judgments, and a human audit, and baseline experiments show BM25 can distinguish original from edited evidence better than dense retrievers.
The paper introduces REVA, a method for compressing retrieval-augmented generation (RAG) prompts by aggregating historical query–document–model interactions into reusable evidence views. REVA mines attention traces from the target generator, maps token-level attention to readable words, aggregates importance across repeated document accesses, and produces budget‑specific plain‑text views that maintain document order and the standard RAG interface. Experiments on four benchmarks with modern LLMs show that REVA improves generation quality by 1.0–5.8 points over existing compressors while reducing compression overhead by 5.3 to 15.6 times and adding less than 40 ms of latency.
arXiv:2607. 10795v1 Announce Type: new Abstract: In open-domain multi-hop question answering (QA), LLM-based search agents offer a promising approach to knowledge-intensive QA by combining retrieval with reasoning.
CoEM introduces a Commit-on-Evidence Memory system that learns when to compress source evidence into compact memory facts while preserving potentially useful excerpts verbatim in a pending set. The system uses a learned policy to decide whether to promote, retain, or discard each pending excerpt as new context arrives, and a frozen verifier ensures only supported facts are committed. Reinforcement learning trains this policy with step-level evidence rewards and final answer rewards, leading to consistent improvements in long-context reasoning, achieving 10.4–11.4 F1 points over the strongest baseline on 6,400-document inputs.
QESChunker introduces a supervised approach to construct static chunk indices for retrieval-augmented generation, using question–evidence pairs to guide the process. It unifies non‑overlapping and overlapping chunking under a single utility‑maximization objective, enabling exact polynomial‑time optimization for non‑overlap and a greedy approximation for overlap. Experiments on two self‑constructed and three public datasets show that both variants outperform existing methods in complete‑evidence recall across most settings.
arXiv:2608. 04569v1 Announce Type: cross Abstract: Hard prompt compression reduces long-context inference cost by independently scoring tokens, sentences, or chunks and retaining the highest-scoring units under a budget.
arXiv:2606. 01542v1 Announce Type: cross Abstract: Chunked-document retrieval is a common component of retrieval-augmented generation (RAG) systems.