What exactly are we measuring when we ask a model to “understand” a task: the next token, the right file, the claim’s true status, or the quality of a multiple-choice item? On one side are papers trying to make pretraining and retrieval more efficient — from lightweight latent multi-token prediction to compact multilingual code-localisation agents using semantic search — and on the other are studies that tighten the measurement itself, including source-aware grading on LongMemEval-500 and LoCoMo-1540, leakage analysis on AVeriTeC and QuanTemp++, and a review of automated QA for multiple-choice assessment items.

AI-Enabled Quality Assurance for Multiple-Choice Assessment Items

For assessment-item QA, separate surface flaw detection from content-sensitive review, because label-level accuracy can hide weak positive-case detection and revision effects need outcome-based measurement.

Generating multiple-choice questions is increasingly scalable, but establishing their assessment quality remains difficult. We present a focused narrative review of automated item-writing flaw detection, revision, psychometric screening, and NLP benchmark auditing. Database searches, citation retrieval, and nominated sources yield fourteen research reports reviewed in full text. We distinguish surface checks from content-sensitive judgments and map a 19-criterion rubric to detection methods and reported evidence. High label-level accuracy often coexists with weak positive case detection, while rubric definitions and reference standards vary. Revision evidence is mixed, and the associations reported in prior work do not establish the effects of repair. We propose evaluating quality assurance as a sequence of independently validated decisions, with criterion-specific reporting, calibrated human review, and outcome-based assessment of revisions.

LightMTP: Lightweight Latent Multi-Token Prediction

If you pretrain MT-adjacent LLMs, latent multi-token prediction can extend supervision to more future tokens with at most 1% extra parameters and without external helper models.

Next-token prediction (NTP) is the standard pretraining objective for large language models, yet it provides an explicit training signal only for the immediate next token, which can lead models to exploit local patterns instead of capturing longer-range structure and ideas. Multi-token prediction (MTP) addresses this by training models to predict several future tokens. However, existing MTP methods often introduce a large number of new parameters with limited improvements in downstream performance. Latent MTP approaches address this efficiency issue by encoding future tokens into a vector representation. However, these approaches usually rely on external helper models for future token encoding. We propose LightMTP, a lightweight, i.e., parameter-efficient, latent MTP approach that bootstraps the future token representations from the model's own hidden states. Our two LightMTP variants extend supervision to more future tokens without requiring the additional computational overhead of conventional MTP nor the external supervision latent MTP normally relies on. LightMTP adds at most 1% extra parameters, retains better performance on general language modeling benchmarks, and achieves similar gains in planning, coding, and reasoning.

Small Agents with Semantic Search: Efficient Multilingual Code Localization

For multilingual code-localisation agents, late-interaction semantic search plus lightweight fine-tuning improves file-finding accuracy and cuts CPU latency by 44.1% while using 29.1% fewer tokens.

Locating relevant files from natural-language requests is a core subtask for agents operating over code repositories. We investigate whether this task can be delegated to compact, specialized models to enable on-device search while reducing the token usage, latency, and inference cost of larger agents. We show that semantic search improves file localization, with gains in accuracy, cross-language transfer, and inference efficiency. To study this setting, we introduce a training framework for file-localization agents built around ColGREP, a local semantic search tool based on late-interaction retrieval models. Our recipe combines weighted supervised fine-tuning on teacher trajectories, assigning turn-level credit based on retrieval outcomes, with reinforcement learning on localization quality. We train three model families with fewer than two billion parameters to formulate search queries, inspect retrieved content, and identify relevant files. On localization tasks derived from SWE-bench Lite and Multi-SWE-bench Flash, ColGREP-equipped agents substantially improve over their base models and outperform corresponding GREP-based agents. In addition to improving localization accuracy, ColGREP reduces mean end-to-end trajectory latency by 44.1% on CPU while using 29.1% fewer tokens, and enables better generalization to programming languages unseen during fine-tuning. These results suggest that compact, tool-specialized localization agents can provide an efficient interface between natural-language requests and large codebases.

MemStrata: 95% and 90.91% Source-Aware Accuracy on LongMemEval-500 and LoCoMo-1540 with a Local Qwen 3.8 27B Q4_K_M Reader

For long-context evaluation, source-aware grading can materially change scores, so a reference-only rubric is not enough when answers must match the full recorded history.

An adequate conversational answer may differ from a short or incomplete benchmark reference. To measure adequacy against the recorded history we prefer source-aware grading, in which the judge checks the reference against the full source before assessing system-blinded answers; original reference-only grading is reported alongside. With a local Qwen 3.8 27B Q4_K_M reader and a 24,000-token evidence ceiling, MemStrata CL1 scores 475/500 (95.0%) on LongMemEval-S and 1,400/1,540 (90.91%) on LoCoMo categories 1-4 under source-aware GPT-5.5 adjudication, against 463/500 (92.6%) and 1,205/1,540 (78.25%) under reference-only grading of the same answers. It preserves a retrieval backbone and adds nonduplicated, dated, speaker-attributed source spans. A same-reader full-history control with about 4.7 times the evidence scores 464/500 reference-only and 470/500 (94.0%) source-aware; neither difference is decisive. Keyword-only selection at the same budget scores 425, and a matched-reader Letta arm 438. On LongMemEval-M, where the packet holds about 1.6% of each history, MemStrata CL1 scores 427/500, with losses concentrated in multi-session and temporal questions. On 300 BEAM-1M questions it outscores dense retrieval, 0.738 to 0.706 (Wilcoxon p = 0.011). A same-seed replay of unchanged requests changed 1.5-2.3% of labels. On identical packets GLM 5.3 flash is non-inferior within 3 points (462 versus 463); Muse Spark 1.3 did not show non-inferiority on 269 questions. None of four pre-registered interventions met all of its registered advancement or feasibility criteria. Signed read-side artifacts support inspection but do not regenerate the private retrieval pipeline. The superiority of source-aware grading to human adjudication is not established, and development exposure, automated-judge dependence and the absence of held-out data preclude an independent-replication or leaderboard claim.

No Hindsight for LLM Fact-Checkers: Measuring Leakage Channels in Misinformation Detection

For fact-checking benchmarks, reconstruct claim-time evidence and watch for parametric leakage, because post-claim retrieval can inflate zero-shot accuracy by 6.3 points.

As automated fact-checking scales on social media, large language model (LLM) verdict scores can look stronger than warranted. One reason is that evaluations mix in information that was not knowable at claim time. Two channels are easy to conflate: outcomes memorized in pre-training and retrieved evidence published after the claim. Yet standard benchmarks rarely separate the two. In this study we measure both channels on AVeriTeC and QuanTemp++ by reconstructing point-in-time evidence conditions and probing for outcome information encoded in model representations. We find substantial evidence of parametric leakage, that can be hidden by the aggregate accuracy, while a simple representation bottleneck reduces this future leakage more efficiently than a mutual-information-based training penalty. We also find that allowing post-claim evidence inflates zero-shot accuracy by 6.3 points in AVeriTeC while the effect is negligible in QuanTemp++, where retrieval provides little post-claim evidence. These results show that misinformation benchmarks can overstate fact-checking performance when they do not account for what information was actually available at claim time.

Papers announced by arXiv on Monday, 5 October 2026, covering submissions from 2 October to 5 October 2026.