Long-form video editing over heterogeneous footage requires agents to coordinate source selection, multimodal analysis, timeline construction, narration and subtitle alignment, rendering, and revision while exposing intermediate state for inspection and repair. We present **Crayotter**, an open-source multimodal multi-agent demo system for prompt-driven long-form video editing. Crayotter organizes production around coverage-aware material preparation, artifact-grounded editing research, and tool-grounded timeline execution. Across these stages, retrieval reports, video analyses, editing blueprints, scheduler events, tool calls, intermediate renders, and final exports are treated as first-class artifacts rather than hidden transient state. The workbench supports local assets, agent-assisted retrieval, progress monitoring, artifact preview, failure diagnosis, interrupted-job resumption, and resource-aware asynchronous execution for long-running workflows. In a 23-theme evaluation, Crayotter achieves the highest human overall score (3.40/5) among the compared systems, with its largest margins in theme alignment, narrative coherence, and editing smoothness. These results show that long-horizon video editing agents can be made traceable, inspectable, and practically controllable through observable production artifacts.

Human evaluation is the gold standard for Natural Language Generation (NLG), but the tooling is fragmented: most platforms serve a single task or a single protocol, and none treat the growing practice of assisting annotators with an LLM as a measurable object of study. These gaps fall hardest on indigenous and under-resourced languages, where paid crowds are scarce, references are unavailable, and data governance matters. We present Aura, an open, self-hostable platform that unifies five evaluation protocols (Likert/DA, pairwise, ranking, error-span, post-editing) across eight task families under one segment schema and one result model, records writing-system and speaker provenance as first-class fields, and runs entirely offline. Its distinguishing feature is a grounded, advisory-only, logged LLM assistant paired with a built-in randomised experiment that estimates the effect of AI assistance on inter-annotator agreement and time-on-task. Aura’s agreement statistics match reference implementations to machine precision. We conducted a human evaluation experiment on three English→Indian-language ranking campaigns (N=9 annotators) which returned a mean usability rating of 4.06/5. Aura is released as open-source software.

Large language model-based generation pipelines commonly transform natural-language requests into structured plans before invoking expensive downstream generators, yet these plans are rarely verified before generation. We present Cinematic Compiler, an interactive system that compiles a short narrative brief into a typed Cinematic Intermediate Representation containing shots, design specifications, entity schedules, and explicit per-shot continuity constraints. A deterministic pre-generation pass checks that every shot carries the required character, costume, prop, and location facts, propagates missing constraints, and rejects contradictory briefs before any generation cost is incurred. At rendering time, the system conditions each shot on a shared character reference image using a frozen commodity image-generation backend, followed by vision-based auditing and reference-anchored repair. Across 15 benchmark briefs, the full constraint pass achieves 100% plan-level constraint coverage, compared with 2.2% for direct prompting, 22.8% for LLM-expanded prompting, and 36.5% for the intermediate representation without propagation. Reference anchoring improves identity-to-reference similarity from 0.605 to 0.714 across 13 briefs, with gains on 12 of 13 briefs, while matched text-only prompting produces no significant improvement. We further show that pairwise cross-shot similarity can reward degenerate near-static generations and therefore report identity preservation jointly with visual dynamic degree. The demonstration allows users to compile, inspect, edit, verify, generate, audit, and repair multi-shot stories through an interactive canvas, while supporting offline replay of stored runs.

Northeast India is home to over 220 indigenous languages, yet none are served by existing agentic AI systems. NE-Agent addresses this gap: a terminal-based agentic system that integrates the NE-Stack, a suite of purpose-built language AI components for Northeast India, into a unified conversational pipeline that dynamically selects between retrieval, translation, and transcription tools based on the user’s query. The system combines language identification across 11 languages (NE-LID), multilingual dense retrieval (NE-Embed, a LaBSE-based model fine-tuned for 10 Northeast Indian languages), Khasi-to-English translation (fine-tuned NLLB), speech transcription across eight languages (NE-ASR), and generation via a locally-hosted open-source LLM, requiring no API keys or cloud infrastructure. A lightweight LLM-based tool router dispatches each query to the appropriate tool, distinguishing NE-Agent from purely retrieval-augmented systems. NE-Agent is open-source, pip-installable (pip install ne-agent), and ships with a multilingual corpus spanning four languages. To our knowledge, it is the first publicly available agentic system built on indigenous Northeast Indian language AI infrastructure: https://github.com/MWirelabs/ne-agent.

We introduce HalluCiteChecker, a toolkit for detecting and verifying hallucinated citations in scientific papers. While AI assistant technologies have transformed the academic writing process, including citation recommendation, they have also led to the emergence of hallucinated citations that do not correspond to any existing work. Such citations not only undermine the credibility of scientific papers but also impose an additional burden on reviewers and authors, who must manually verify their validity during the review process. In this study, we formalize hallucinated citation detection as an NLP task and provide a corresponding toolkit as a practical foundation for addressing this problem. Our package is lightweight and can perform verification in seconds on a standard laptop. It can also be executed entirely offline and runs efficiently using only CPUs. We hope that HalluCiteChecker will help reduce reviewer workload and support organizers by enabling systematic pre-review and publication checks. Our code is released under the Apache 2.0 license on GitHub and is distributed as an installable package via PyPI. A demonstration video is available on YouTube.

Flagship language models appear saturated on benchmarks like MMLU (Hendrycks et al., 2021), scoring above 90% - yet benchmarks test only what the experimenter thought to ask, the availability bias of fixed question sets. LLMPEDIA makes this bias measurable and browsable. We recursively materialized ∼1.3M articles from three model families’ parametric memory (GPT-5-mini, DeepSeek-V3.2, Llama-3.3-70B) without retrieval, then audited a stratified sample of atomic claims against Wikipedia and a curated web stack, coloring every claim supported, refuted, or insufficient (Saeed and Razniewski, 2026). On a uniform random sample the true rate is 68.4% - more than 21 pp below MMLU - with 30.5% of claims insufficient: assertions no benchmark probes and the world’s largest encyclopedia cannot adjudicate - long-tail knowledge or plausible hallucination, the evidence cannot tell - extending to free text the coverage gap GPTKB established for triples (Hu et al., 2025). The resulting live, open encyclopedia lets visitors inspect this frontier one claim at a time through five one-click views - link-traversal exploration, claim-level factuality, cross-model and political-persona comparison, and a guided topic drill-down - each page, claim, and verdict at a stable URL. LLMPEDIA is live at https://llmpedia.net; see the demonstration video.

Evaluating open-ended generation remains non-trivial because quality varies across style, correctness, grounding, and safety. Moreover, a fixed prompt-response abstraction is insufficient for a reusable evaluation system as deployed applications increasingly expose multi-turn dialogues, retrieved evidence, and tool-calling traces. We present Pollux, a universal criteria-based evaluation framework centered on Pollux Judge, a compact Qwen3-4B-based LLM-as-a-Judge that evaluates such heterogeneous interactions through free-form user-defined criteria, scales, and output formats. Pollux follows the criteria methodology represented in the Pollux Benchmark (35 task groups, 2,115 expert-authored prompts), using it as a human-grounded resource for judge validation. The judge is trained with GRPO on Russian- and English-language data and is designed for bilingual deployment through a language-independent interaction schema. On the held-out expert-labeled test set, Pollux reaches a Spearman correlation of 0.744, exceeding every evaluated general-purpose baseline, including models with up to 1.1T parameters, while remaining suitable for local or OpenAI-compatible serving.

System demonstration video: https://drive.google.com/file/d/1PF6L-RcrS24JVUdaVRi01j237yTdftEd/

This paper proposes AI Tour Meeting, a group travel planning framework powered by multiple Large Language Model (LLM)-based agents. The agents are instantiated with distinct personas and collaboratively seek an itinerary that satisfies their constraints and preferences through natural language discussion.

The framework enables easy and flexible orchestration of such discussions by providing interfaces for configuring agent personas, discussion workflows, monitoring, and LLM deployment. Its primary use case is a simulation tool for analyzing the behavior of multiple LLM agents during tour planning discussions.

This paper demonstrates the utility of the framework by presenting system validation and several analytical results obtained using the framework.

The logit lens is a standard mechanistic interpretability tool for inspecting a Transformer’s predictions across layers. However, existing logit lens tools typically focus on individual checkpoints rather than interactive exploration of a model’s *training history*.

In this study, we introduce **LensLapse**, an open-source web application for interactive exploration of a model’s training history through token-level interpretability. Users can scrub a slider across public checkpoints from five model families and three multilingual checkpoint suites and observe how per-layer, per-position logit lens predictions evolve from frequency bias to factual knowledge.

For browser-compatible models, arbitrary user prompts are probed *live in the browser* via local inference over per-checkpoint model pairs. No server, no API key, no hosting cost, and prompts never leave the user’s device.

Moreover, **LensLapse** supports local models. A single ‘pip install‘ enables users to launch the application and explore their own models and training checkpoints without any additional web setup.

LensLapse is MIT-licensed and permanently hosted at iamtatsuki05.github.io/lenslapse.

- GitHub: https://github.com/iamtatsuki05/lenslapse - PyPI: https://pypi.org/project/lenslapse/ - docs: https://iamtatsuki05.github.io/lenslapse/docs/ - YouTube: https://youtu.be/OBnvG9Ru3j8

Large language models can transform project documents into software requirements, but direct generation may introduce unsupported claims and hidden inconsistencies. We present ReqLens, an interactive multi-agent NLP system that extracts candidate requirements from mixed project documents, links them to source evidence, exposes graph-based dependencies and conflicts, and generates a Software Requirements Specification (SRS) through a human-in-the-loop workflow. We evaluate ReqLens on 13 projects using direct-generation and modular baselines, as well as component ablations. ReqLens achieves the strongest overall balance across unsupported-requirement rejection, valid-requirement retention, and seeded-defect filtering. A formative study with four domain experts provided qualitative feedback on the system’s interface and review workflow. ReqLens is accompanied by code, data, an installable package, and a narrated screencast.

Final-answer accuracy hides where a mathematical solution first becomes inconsistent and whether a later correction merely masks that error. We present StateScope, an interactive system for detecting, localizing, and repairing solution-state drift: divergence between a solver’s evolving state and a symbolically verified reference trajectory. Under a controlled comparison, two solvers receive the same Computer Algebra System (CAS)-derived operation schedule and typed arguments. We provide two systems, 1) System C records the model’s textual state claim, 2) System D re-derives state in an external typed ledger. A SymPy oracle checks every transition and explains component-level differences without an LLM judge. Users can step through a run, edit an operation at any retained step, directly repair System C’s claimed state, continue the model, and repeatedly branch from earlier interventions. Our paired evaluation covers 250 problems and twelve open and proprietary models, and a human study measures diagnostic accuracy and time. StateScope targets reasoning researchers, model developers, and educators who need to inspect and manipulate more than a final answer.