While prior work has highlighted that systematic nation-level bias in large language models (LLMs) can pose operational risks for international relations (IR) applications, many existing evaluations still lack a clearly specified unbiased reference (ground truth), limiting fully quantitative and cross-setting measurement. Using an expanded corpus of 722 official UN Security Council (UNSC) records, we measure nation-level bias with five complementary tests, thereby extending prior approaches to cover a broader set of bias expression channels: opinion ranking (WOW), association sorting (AK+-), vote simulation (PVS), opposition attribution (F-Oppose), and framing (Headliner). To enable grounded bias evaluation, we explicitly define an unbiased status for each test; in particular, for WOW and AK+-where ground truth is difficult to specify, we propose an action-based pseudo unbiased status derived from observable voting behavior to construct a quantifiable baseline. Benchmarking 4 LLMs (GPT-4o-mini, Llama 3.3-70B, Mistral-22B-Small, and Qwen3-30B) across the five permanent members of the UNSC, we find a test-agnostic macro-pattern: relative to the reference baselines defined for each test, model outputs lean toward the opposition (red) side for China and Russia and toward the supportive (blue) side for the U.K., the U.S., and France. These results underscore the need to audit IR-facing LLM deployments with grounded, multi-test benchmarks tied to explicit reference points.

Document weighting is an underspecified bottleneck in black-box retrieval-augmented generation. Training-free systems inherit answer-aggregation weights from retrieval similarity, yet this proxy has not been systematically compared with evidence-based alternatives. We propose a unified two-parameter framework that generalizes REPLUG with an evidence term; it is training-free, black-box, and needs no large language model (LLM) calls beyond voting. Similarity-only (λ=0) and evidence-only (λ→∞) weighting are its limiting cases, so one interpolation parameter separates the retrieval prior from the evidence signal. We evaluate across 3 LLMs (7B–70B), 3 open-domain question-answering datasets (31,861 queries), and 3 off-the-shelf evidence types. Cross-encoder and embedding-stability evidence improve exact match in all 9 LLM–dataset cells, whereas natural language inference (NLI) fails as a weighting signal, to our knowledge the first systematic evidence of this failure mode. The similarity prior is largely redundant on NQ-Open and PopQA but helps on TriviaQA, where retriever scores track gold documents better; the pattern replicates under the E5 retriever, with smaller gains concentrated on long-tail PopQA. One default transfers across datasets and scales with minimal loss; calibration depends on signal type, not scale. Practically: prefer cross-encoder evidence, avoid NLI, and use vote mass for selective prediction; weighting itself adds sub-millisecond overhead.

Similarity-based retrieval is the dominant rule for conditioning large language models (LLMs) in in-context learning, retrieval-augmented generation, and prompt-based time series forecasting. The rule concentrates on near-duplicate candidates, an issue that has motivated diversity-aware retrieval but remains unexamined in other retrieval-conditioned pipelines. We study this issue using prompt-based time series forecasting as a test bed, where a learned prompt pool is retrieved by similarity. Dominant methods in this setting retrieve top-K entries by cosine similarity without redundancy control, producing a bias toward dominant temporal patterns while overlooking rare but informative events. We propose CASP-LLM, a coverage-aware semantic prompting framework that addresses this prompt selection bias by combining usage-tracking and saturating-gate techniques into a coverage regularizer that adds no learnable parameters. On six long-term benchmarks and the M4 short-term benchmark, CASP-LLM matches or improves on similarity-based LLM forecasters on most dataset-horizon settings, with the exceptions of Electricity, M4-Monthly, and the few-shot long-horizon setting. A controlled study locates the failure mode at the cross-batch usage level rather than per-retrieval redundancy: within-retrieval diversification such as MMR does not help, whereas regularizing anchor usage across training does.

Time series forecasting underpins critical decision-making across diverse domains. While large language models (LLMs) offer promising reasoning capabilities, existing LLM-based time series forecasting approaches either reduce them to numerical predictors that bypass their strengths, or allow direct forecast generation that destabilizes predictions in non-stationary settings. We introduce CTRL, a framework that decouples semantic reasoning from quantitative prediction. A frozen backbone generates base forecasts, while specialized LLM agents function as controllers that analyze backbone prediction errors through decomposed trend, seasonal, and irregular components, grounding reasoning in interpretable temporal structure. Each agent outputs compact control signals that a lightweight residual decoder translates into forecast corrections. CTRL incorporates label-free test-time adaptation that detects distribution shift from input statistics alone and readapts control signals with only 3–24 LLM calls via caching. CTRL is explicitly designed to improve robustness under non-stationary temporal dynamics and distribution shift, while remaining competitive on highly stationary time series where adaptive correction provides limited additional benefit.

Current retrieval-augmented generation (RAG) methods struggle with complex multi-hop reasoning, relying on unstructured semantic matching that lacks the logical structure needed to systematically guide retrieval. We introduce Programmatic Retrieval Optimization with Generative Reasoning and Augmented Multi-queries (PROGRAM), a novel framework that elevates retrieval to structured, program-guided reasoning. PROGRAM treats retrieval as execution of specific program types, such as logical, temporal, causal, and so forth, through three stages of ’Program-Type Selection’ with dual-metric optimization, ’Iterative Active Program Pruning’ with evidence accumulation, and ’Final Answer Generation’ with reranking. Evaluated on five benchmarks including HotPotQA, 2WikiMultihopQA, ARC-Challenge, MMLU-Pro, and MedQA with various LLMs, PROGRAM achieves state-of-the-art performance with up to 24% relative improvement on HotPotQA and 13.2% on MedQA over strong baselines including FLARE, ProbTree and Self-RAG.

Open-world logical entailment requires deciding whether a set of premises entails a conclusion, entails its negation, or leaves it undecided. Many existing methods are deductive in spirit, searching for a proof and reading the label from whether that proof succeeds. When implemented through LLM-based proof search, this makes the prediction depend on a single unstable trajectory, and a failed or partial search is easily mismapped to the wrong label. We propose ABED (Assumption-Based Bidirectional Entailment Diagnosis), a training-free inference-time framework that reframes entailment classification as missing assumption diagnosis. Instead of asking only whether a conclusion can be proved, ABED asks what assumption is missing for it to hold, and classifies that gap as already closed by the premises (Null), contradicted by them (Contra), or left open (Admissible). It applies this diagnosis to both a conclusion and its negation, and determines the label from the resulting dual gap pair. Across FOLIO, ProofWriter, ProntoQA, and LogicNLI, ABED consistently outperforms prompting-based, symbolic, solver-based, and deductive baselines, improving over the strongest baseline on each dataset by 3.4, 5.3, 0.8, and 4.4 points with GPT-4.1.

Large language models with retrieval-augmented generation systematically underutilize evidence in the middle of multi-document contexts, a phenomenon known as Lost in the Middle. Existing mitigation strategies, such as document reordering and attention steering, can reduce boundary bias but remain brittle in noisy settings, where inaccurate relevance estimation and spurious documents still obscure useful evidence in central positions. We propose DuQUAD, a multi-document QA framework that mitigates positional bias through dual-view reasoning and quality-aware evidence pruning. DuQUAD combines a Local Agent guided by a Structural Fusion Score with a Global Agent that reasons over the full document set, providing both targeted evidence recovery and broad contextual understanding. The framework further filters candidate evidence using explicit quality scoring and refines it at the sentence level to reduce noise. DuQUAD consistently outperforms strong Lost-in-the-Middle mitigation baselines, including recent multi-agent and context optimization methods, improving accuracy by 3.9% on average over the strongest competing baseline and gold evidence recall by 9.4%, while substantially reducing computational cost compared to existing multi-agent frameworks.

Knowledge-intensive queries require accurate answers that are explicitly grounded in retrieved evidence. However, existing retrieval-augmented generation (RAG) approaches often struggle with query complexity, suffer from propagated reasoning errors, or rely on incomplete or noisy retrieval, limiting their effectiveness. To address these limitations, we introduce UniRAG, a unified RAG framework that integrates entity-grounded query decomposition, break-down reasoning, and iterative query rewriting. Specifically, UniRAG decomposes queries into semantically coherent sub-queries, explicitly verifies retrieved sub-facts through a dedicated reasoning module, and adaptively refines queries based on identified knowledge gaps, significantly improving answer completeness and reliability. Extensive benchmark evaluations on complex question-answering datasets, including multi-hop HotPotQA and 2WikiMultihopQA, biomedical MedMCQA and MedQA, and fact-verification FEVER and SciFact, demonstrate that UniRAG consistently achieves performance improvements across various state-of-the-art LLMs, such as LLaMA-3.1-8B, GPT-3.5-Turbo, and Gemini-1.5-Flash.