Game media and interactive narrative

Cinematics, playable stories, branching narrative, worlds and media assets

• articleHow do prompted frontier LLMs compare with fine-tuned small models at next-state prediction in text environments (including TextWorld), and when did predicted rollouts stay consistent with the real environment?
  Corrected 2026-10-24 - see C3-B1131-01; the original statements below are kept for history and are superseded where the correction says so. Li et al. (arXiv 2512.18832 v2) measure exact-match (EM) next-state accuracy in five text environments. Prompted models (zero-shot or 3-shot) reached 0.00-17.70% zero-shot and 11.43-49.12% 3-shot on TextWorld, the narrative-game engine; fine-tuned Qwen2.5-7B and Llama-3.1-8B (trained on 40K GPT-4o trajectories per environment) reached about 70.5%. On ALFWorld/SciWorld fine-tuned models reach 99%+ but zero-shot frontier models 35-65%. Consistency of rollouts (success when replaying world-model actions in the real environment divided by real success) was 0.91-0.96 in the three structured environments for the fine-tuned Qwen world model and 0.67 on WebShop. Fine-tuned models are not "the LLM itself" prompted; the authors say so by contrasting prompting with SFT.
• articleGame writers use 'AI director', 'storyteller', 'emergent narrative' and 'drama manager' loosely; what does each actually control in a shipped game, and are they the same thing?
  Corrected 2026-10-24 - see C3-B1130-01; the original statements below are kept for history and are superseded where the correction says so. In the first-person sources read, the shipped 'directors' control less than the names suggest. Left 4 Dead's AI Director modulates how many threats spawn from an estimate of player intensity (pacing, not plot or difficulty); RimWorld's 'AI storyteller' chooses which random incidents arrive, with one preset aiming at a tension curve; Dwarf Fortress has no director at all and relies on autonomous simulated characters. 'Emergent narrative' is the academic umbrella term for stories arising bottom-up from such simulations, and 'drama management' is Ryan's name for the different idea of an omniscient agent intervening to enforce story structure. The sources do not use the words consistently, so an agent should ask what the controller touches (spawns, incidents, nothing, story structure) rather than trust the label.
• articleWhat happens inside a streaming video model when the prompt changes mid-rollout, which cache tricks do the papers use, and which failure modes do they report?
  Corrected 2026-10-24 - see C3-B1086-01; the original statements below are kept for history and are superseded where the correction says so. Four papers handle a mid-rollout prompt change by editing the attention cache. LongLive recomputes cached key-value states with the new prompt (KV recache) and reports the trade-off: with no cache the new prompt is followed but transitions are abrupt; with a plain cache visuals are smooth but the new prompt is delayed or ignored. Zing-0.5 refreshes only the text K/V and pins the first frame after the switch; GameCraft-2 uses a prompt recaching mechanism with a sink token; WorldSonus replaces its cross-attention cache at the next 100 ms chunk boundary. MaineCoon gates the next prompt on whether the current spoken line has finished. Reported failure modes: prompt inertia, abrupt transitions, adherence decay across switches, drift beyond about 500 frames and no persistent entity state.
• articleWhat do MASS (authoritative learned state) and WorldCast (distributed clients exchanging state) show about moving the shared state out of the video generator, and what do their evaluation protocols leave untested?
  Corrected 2026-10-24 - see C3-B1085-01; the original statements below are kept for history and are superseded where the correction says so. MASS shows, on a matched Snake benchmark at 128x128 and 128 ticks, that a typed learned state recovers more objects (parser 0.764 versus at most 0.128 for video baselines) and has zero cross-view disagreement by construction, but its structural-validity score is worse than independent per-view baselines and its local prediction decays when other players inputs are missing. WorldCast shows on Counter-Strike 2 that exchanging camera-aligned player state raises the rate at which other players are drawn (0.817 versus 0.056 and 0.074 for retrained baselines), but evaluates lockstep clients without network delay or faults and with ground-truth states for most whole-round results. Neither releases code or weights that were found.
• articleI need to pace tension in a systemic game or simulation: what do designers actually do (build-up, peak, relax), what goes wrong, and which parts of a pacing controller are documented versus my own proposal?
  Corrected 2026-10-24 - see C3-B1070-02; the original statements below are kept for history and are superseded where the correction says so. Documented practice (Valve, Digital Extremes): measure a crude per-player intensity from damage and kills, cycle build-up, peak, hold-off and relax states, keep set-piece bosses outside the throttle, telegraph big events, plan density peaks rather than uniform density, and let designers override the director in special modes. Documented failure modes: fatigue from constant combat and boredom from lulls, memorized static triggers, events the player cannot perceive, raw event streams that are not stories, narrative that simply does not emerge, boring output from thin simulations, and unfair random clusters. Players and communities, not the games, do most of the retelling. Everything about how to combine these into one controller, and how to test it, is the worker's proposal and is labelled below; no source tests a complete recipe.
• articleWhich open-weight interactive or action-conditioned world models were released after Oasis and DIAMOND (Matrix-Game 1.0-4.0, Hunyuan-GameCraft, LingBot-World 1 and 2), what is actually released for each, and what licence is attached to which part?
  Corrected 2026-10-24 - see C3-B1075-01; the original statements below are kept for history and are superseded where the correction says so. Eight released model lines with public weights were found: Matrix-Game 1.0 (MIT), 2.0 (MIT), 3.0 (Apache-2.0, 5B base and distilled; a 28B model is announced as 'soon'), 4.0 (Apache-2.0; scene checkpoints for three games, released 2026-10-15), Hunyuan-GameCraft-1.0 (Tencent Hunyuan Community License with territory exclusions), LingBot-World 1 (Apache-2.0) and LingBot-World 2.0/Infinity (CC BY-NC-SA 4.0, non-commercial). Licences differ by version inside one project and by component (code, weights, game assets, bundled third-party models). 'Open weights' here means files present on Hugging Face at the pinned revision with a licence text; nothing was downloaded or run.
• articleFor each of these open interactive world models, what do the papers and READMEs say about the action space, output resolution, frame rate, context or memory, and hardware, and which cells are measured, claimed or absent?
  Corrected 2026-10-24 - see C3-B981-01; the original statements below are kept for history and are superseded where the correction says so. The numbers are not comparable across systems because they mean different things: Matrix-Game 2.0's 25 FPS is generation speed on one H100 at 352x640; Hunyuan-GameCraft's 25 fps is the video frame rate while its own table lists inference speed of 0.25 FPS (6.6 with distillation) and the text says latency under 5 s per action; LingBot-World 1 claims under 1 s latency at 16 fps; LingBot 2.0 claims 720p at 60 fps without a hardware statement in the text read; Matrix-Game 4.0 takes player input in a game engine and renders low-poly frames, reporting 480p 60 FPS on an RTX 5090 with 112 ms key-press-to-display latency. Memory is explicit only in Matrix-Game 3.0/4.0 (README/report statements) and is stated as absent or emergent in Matrix-Game 2.0 and LingBot-World.
• articleHow was the multi-player training data in these papers obtained (bots, simulators, game recordings, single-view video), and what does each source imply about generalising to human play and about rights?
  Corrected 2026-10-24 - see C3-B1057-01; the original statements below are kept for history and are superseded where the correction says so. Five data routes appear: scripted Mineflayer bots in Minecraft (Solaris, Gamma-World), publicly available bots in Rocket League (MIRA), a driving simulator (ShareVerse, CARLA), recordings of a co-op game (MultiWorld, It Takes Two, natural play) and of Counter-Strike 2 matches (WorldCast, OpenCS2; whether players were human is not stated in the passages read), and camera-trajectory unrolling of single-view video (MetaWorld). Authors flag bot-only data as a generalisation gap (Solaris, MIRA). Only MIRA states a publisher permission for its gameplay recordings in the text read.
• articleHow do Niagara Fluids differ from an offline Houdini or Blender sim: what grids and solvers do they use, what is 2D versus 3D versus FLIP, and what does Epic say you should bake?
  Problem: deciding whether a real-time fluid sim in Unreal can stand in for an offline pyro or liquid sim. Conditions: Unreal Engine 5 Niagara Fluids plugin. Method: gases run on grids holding density, temperature and velocity per cell; liquids use a hybrid FLIP (particles plus a grid velocity solve); pressure solve iterations trade accuracy for cost. Epic names simulator types: 2D Gas (much faster, suited to games, lacks complex 3D turbulence), 3D Gas (depth and richer flow at higher memory and GPU cost, for hero effects and cinematics), 2D FLIP (splashes, not pools), Shallow Water (height field for pools and wakes) and 3D FLIP (waves, rivers; computationally expensive). Results can be baked to flipbook textures for cheaper playback, and colliders are static meshes, Geometry Collections or depth maps. Example: a 3D gas torch or explosion baked to a flipbook for a mobile-friendly asset. Failure modes: Epic warns that fluid scenes can be graphically intensive and may crash the GPU on ...
• articleHow does Wave Function Collapse generate a tilemap or bitmap, what does it guarantee, and when does it fail (contradiction)?
  Problem: Generate locally-consistent tilemaps or textures from an example. Conditions: input is a small bitmap (overlapping model, NxN patterns, typically N=3) or a tile list with adjacency rules (simple tiled model). Method: count NxN patterns (optionally with rotations and reflections); start with every output cell in an all-allowed superposition; loop: observe (pick the cell with the lowest non-zero Shannon entropy and collapse it using the pattern frequencies), then propagate the removed options through neighbours; stop when everything is collapsed. Guarantees in the README: (C1) output contains only NxN patterns present in the input; (weak C2) pattern distribution is similar to the input. Failure modes: if propagation zeroes every option in a cell, the run hits a contradiction and returns nothing; the README says deciding whether other valid bitmaps exist is NP-hard, so no fast always-finishing solver exists, yet contradictions are rare in practice (author's statement, not meas...
• articleHow should an agent track narrative state across scenes so that scenes can be reordered, resumed and checked, given what storylet and Twine engines record?
  Problem. Agents drift and contradict earlier scenes. Conditions. Multi-scene or interactive scripts written with or for tools like ink, Yarn, SugarCube and Harlowe. Synthesis (inference, built from the cited findings, not a source recommendation). (1) Separate durable facts (what the character knows, relationships, inventory) from scratch values, as SugarCube separates story and temporary variables. (2) Record visit history per scene as an ordered list plus counts (Harlowe history/visits; ink and Yarn counts in C3-B542-01/B543-01), and derive 'has seen' checks from it rather than separate flags. (3) Model each scene as a unit with preconditions, content and effects (storylet anatomy) so availability and consequences are explicit and testable. (4) Keep deterministic state outside the model when conditions are LLM-judged. (5) Use mock visits or fixtures to test branches. Failure modes. Stale flags, expired history, unlogged effects. Limitations. This is a design synthesis; none of the...
• articleHow does Harlowe track visits and history, and when should I use (visited:) instead of a flag variable?
  Problem. Writers duplicate engine state in hand-made flag variables. Conditions. Twine stories using Harlowe 3.3.x. Method. (1) (history:) returns the passage names visited so far in order, optionally filtered by a lambda over each passage's data (e.g. tags), so counts per tag come from its length. (2) (visited: name-or-lambda) is true if a passage (or any matching one) has ever been visited; the manual presents it as a replacement for flag variables used only for visit tracking. (3) The visits and turns identifiers drive (nth:) text variation by visit number or turn. (4) Passages reached through (redirect:) count as visited. (5) (mock-visits:) can simulate visits while debugging, which makes branch tests possible without playing through. Failure modes. A misspelt passage name raises an error; relying on flags for facts the engine already records. Limitations. Does not cover save slots or version loading (C2-B08-F-05). Behaviour read from the manual, not run. Version/date. Harlowe 3...
• articleIn SugarCube, which story state is saved and undoable, which is temporary, and how do I check whether the player has ever visited a passage?
  Problem. Narrative state must survive saves and undo but not every scratch value should. Conditions. Twine stories using SugarCube 2. Method. (1) Story variables ($name) are part of the story history and live for the playthrough; temporary variables (_name) are not part of history and live for one moment/turn, suited to loop counters. (2) The history is a list of moments; the active moment holds the variables. (3) For 'has the player ever been here', use hasVisited(...) or State.hasPlayed(...) which include expired moments; State.has() only checks the past in-play history so it can miss passages whose moments expired. (4) lastVisited(...) returns turns since a visit, -1 never. (5) Use previous() for the last non-active passage. Example. <<if hasVisited("Bar")>> for a one-time reaction. Failure modes. Using State.has() for permanent flags when history is capped; putting durable facts in temporary variables. Limitations. Extends the ink and Yarn visit semantics already catalogued (C3-...
• articleWhat viewing-comfort practices does Apple's HIG give for video in visionOS?
  Problem. Immersive video can be uncomfortable when forced on viewers. Conditions. visionOS apps that play video they may not control. Method (HIG). Let people choose when playback starts; use a small playback window that people can resize; keep their surroundings visible during playback; in fully immersive experiences do not let virtual content obscure playback or transport controls; do not start a fully immersive video automatically; provide a thumbnail track (160 px wide thumbnails) for scrubbing; keep inline video 2D and do not expand it to fill a window. Failure modes. Auto-starting immersive playback; occluding controls. Limitations. These are app-level comfort practices, not disparity or convergence limits. Version/date. HIG documentation JSON fetched 2026-10-23.
• articleWhat rules does Don Carson's environmental-storytelling essay give for a consistent, readable story world?
  Problem. Spaces that look good but tell nothing, or break their own rules. Conditions. Game or themed-environment design (essay from March 1, 2000, drawing on theme-park practice). Method. (1) Define a big-picture story as a set of rules guiding design and team; never break those rules. (2) Answer 'Where am I?' within the first 15 seconds, then 'What is my relationship to this place?' through clues rather than text dumps. (3) Use cause-and-effect vignettes: staged areas that let the player conclude what happened (broken doors, a crashed vehicle, charred remains), also to show time passing or player-caused changes. (4) Follow 'bread crumb' trails left by a fictional predecessor. (5) Give familiar anchors in alien environments. Failure modes. Adding elements just because they are cool; alienating all-alien levels. Limitations. Craft essay with examples from 1990s games and attractions, not empirical evidence. Version/date. Page dated March 1, 2000; fetched 2026-10-23.
• articleHow should detail, light and props be placed to direct attention in a story space, per Carson?
  Problem. Over-detailed, evenly lit spaces hide story cues. Conditions. 3D level or set design. Method. (1) Use contrast: a large room with one shaft of light on a solitary prop beats a room full of detailed elements; make other props supporting players to the important story element. (2) Use fewer, quieter textures and save the most decorative element for the end of a corridor to pull the viewer on. (3) Avoid flooding a level with light, which flattens depth; let large areas fall into shadow while not hiding vital game elements. (4) Do not offer too many choices at once. (5) Treat texture maps as stage canvas, not wallpaper. Failure modes. Lighting that reveals every nook; cluttered high-frequency textures. Limitations. Craft guidance from an era of limited rendering; principles are compositional. Version/date. Essay dated March 1, 2000; fetched 2026-10-23.
• articleWhat four ways does Henry Jenkins say game spaces carry story (evocative, enacted, embedded, emergent), and how do they combine?
  Problem. The ludology/narratology dispute stalls practical design. Conditions. Designing spatial or open-world narrative. Method. Jenkins proposes game designers as narrative architects. Environmental storytelling sets the preconditions for narrative in at least one of four ways: spaces evoke pre-existing narrative associations (genre/story worlds), act as a staging ground for enacted events, embed narrative information in the mise-en-scene for players to discover, or supply resources for emergent narratives (his example: The Sims). Detective-style design: a loosely structured story driven by the player's exploration plus a pre-structured one embedded in the space awaiting discovery; mixing enacted and embedded elements balances interactivity flexibility with coherence of a pre-authored narrative. Failure modes. Treating a space as empty of narrative because it lacks cutscenes. Limitations. Essay from the early 2000s with game examples (Myst, Half-Life, The Sims); fetched text carri...
• articleHow should an AI world or scene generator place story cues, combining Carson's and Jenkins's craft?
  Problem. Generated scenes are often visually dense but narratively empty or contradictory. Synthesis (inference, built from findings 11 to 13, not from a source recommendation). (1) Start from a rules statement for the world and check every generated element against it (Carson: do not break the rules). (2) Make the first view answer where the player is within about 15 seconds, using one dominant landmark and a few genre-evoking cues (Jenkins: evocative spaces). (3) Place a small number of cause-and-effect vignettes with clear traces of a prior event, each with a single highlighted focal prop (Carson: vignettes, single shaft of light). (4) Tag each cue as evocative, enacted, embedded or emergent and make sure embedded cues are consistent with the story ledger and discoverable. (5) Reserve decoration to lead the eye toward the next space; avoid evenly lit clutter. (6) Keep a list of cues so later scenes can alter them to show time passing. Failure modes. Random props that contradict t...
• articleDo Wwise's music transition rules treat a segment's Exit Cue as a legal 'Next Grid' exit point, per any retrievable source?
  The only retrievable source is a user feature request (Wwise 2021.1.6), not documentation. It reports that 'Exit source at Next Grid' does not treat a segment's Exit Cue as an exit point even when the cue falls on a grid line, so transitions can be delayed to the next grid point; users work around it with custom cues on the grid and Next Cue exits. The vendor reply only thanks the poster, so the behaviour is neither confirmed nor denied, and later versions are unknown.
• articleDo Unity Timeline and Godot document any editorial exchange (EDL, XML, OTIO, USD) for cinematics?
  Within the documents inspected, neither engine documents editorial interchange for cinematics. Unity Timeline's 1.8 manual has no export, EDL, XML, OTIO or import topic, and says its frame rate is only visual. The godot-docs tree has no editorial or OTIO pages. OpenTimelineIO lists adapters for NLE and DCC formats (including Maya sequencer and Kdenlive) but none for Unity, Godot or USD. This is a bounded negative; third-party plugins may exist.
• articleHow does Suspenser choose which story events to tell to raise suspense, and what did its human evaluation show?
  Partially verified: the curator re-read none of the cited sources; S7, S8 rest on the producer's pinned reading. Suspenser (Cheong and Young) chooses a story's content from a planned fabula so that readers can infer few complete plans for the protagonist's goal: it keeps the causally important events as a skeleton, then swaps in events that threaten the protagonist's goals when that lowers the number of solutions the modeled reader can find. In a 98-person study with three fabulas, Suspenser's selections were rated as suspenseful as one human author's high-suspense selections and more than the same author's low-suspense selections, with moderate absolute ratings. The authors say the design cannot say which component caused the effect.
• articleDo open papers define standard metrics for branching story graphs (reconvergence, reuse, diversity), and what values do they report?
  Partially verified: the curator re-read none of the cited sources; S16, S17, S18 rest on the producer's pinned reading. In the three open arXiv papers read, no standard reconvergence or choice-density metric for authored branching stories is defined. MACHIAVELLI reports corpus scale only (134 games, 572,322 scenarios, 4,559 achievements). CHADPOD builds 1,462 decision-point classification tasks from CYOA graphs. Forking Garden (2026, LLM story-graph generator) defines an entity-overlap edge score and reports an entity-diversity ratio of 0.678 versus 0.261 for a baseline, a paper result on generated graphs, not on authored works.
• articleHow do I extract a graph from a finished branching work, what open graph data exists, and what choice-density measures have been used?
  Partially verified: the curator re-read none of the cited sources; S6, S7, S5, S8, S9 rest on the producer's pinned reading. Three first-person web sources: a 2017 post gives an edge-per-instruction extraction recipe (look at each page once, write an edge, label endings, render with Graphviz) and notes interest in reconvergence and cycles; the Transverse Reading Gallery states it holds 834 network graphs for 2,569 books with DOT and TGF files and edge lists; Swinehart's project classifies pages as branching, story or ending and plots page-type shares across 12 books, reporting a decline in endings and choices over time (qualitative reading of charts; no numeric table in the text read). Together they give an extraction method, open graph files and choice-density framing.
• articleWhich named structure patterns do practitioners use for choice-based stories, and what does each one feel like to play and need in state?
  Partially verified: the curator re-read none of the cited sources; S1 rests on the producer's pinned reading. Ashwell's 2015 post (first-person practitioner/critic, one author) names eight non-exclusive patterns and for each gives structure, state-tracking need, effects on the player and examples: Time Cave, Gauntlet, Branch and Bottleneck, Quest, Open Map, Sorting Hat, Floating Modules, Loop and Grow (plus spoke-and-hub as a variant). He states the categories are not discrete and the example diagrams are simpler than real works. Pinned raw bytes were read here; the catalogue's earlier observation was a tool-text read with no retained bytes.
• articleBeginner: has any film, advertising, series or game job reported using an action-conditioned interactive world model (Genie 3 / Project Genie, Odyssey-2, Runway GWM-1, NVIDIA Cosmos) with a measured outcome?
  Partially verified: the curator re-read none of the cited sources; S7, S8 rest on the producer's pinned reading. Negative result, bounded: no production case with a measured outcome was found for interactive/action-conditioned world models. The two company pages read first-hand (Odyssey-2, Runway GWM-1) list media uses as expected or listed applications and name no customer or result. Search found nothing further, and two sources (Project Genie blog, an ACM paper) could not be read. This is absence within the pages read plus discovery searches, not proof of absence.
• articleDo independent creators' micro-shorts, episodic and VR/game prototypes built on AI-generated worlds report any measured outcome (time, rework, shots used, ratings)?
  Partially verified: the curator re-read none of the cited sources; S3, S4, S5 rest on the producer's pinned reading. Four creator projects (Alarm, Cryptid Worlds, Memory House, Splat World) are documented on World Labs pages dated 2025-11-12. They describe workflows and qualitative outcomes (consistency, 'within hours', platform stability) but no measured time, cost, shots used, rework or quality ratings. Evidence class: vendor-authored case studies quoting the creators.
• articleWhat checks or playtest measures do the open sources propose or use for player agency in interactive narrative?
  Partially verified: the curator re-read none of the cited sources; S7, S6, S3, S1, S4 rest on the producer's pinned reading. No inspected source supplies a validated agency instrument. What exists: Thue et al. asked 90 students to rate their experience after fixed or adaptive stories and tested whether adaptive stories felt more influential, with weak overall support and stronger results in filtered subgroups; Wardrip-Fruin et al. report Dow's study of three Facade versions using interviews, finding that greater presence and realism reduced agency by raising expectations; Facade's authors give self-estimated local-agency percentages; Mateas's structural model suggests an analytic check. Roberts and Isbell list measurability as a desideratum.
• articleInstead of trusting an LLM to simulate each step, can it write an explicit rule model (PDDL) that a validator and planner then use, and how many errors did the LLM make doing so?
  Partially verified: the curator re-read none of the cited sources; S3 rests on the producer's pinned reading. Guan et al. (NeurIPS 2023, arXiv 2305.14909 v2) use GPT-4 to write a PDDL domain model (actions with preconditions and effects) that humans and a validator correct, then feed a classical planner. GPT-4 made 53 errors on a 22-action Household model, 2 on Logistics (6 actions) and 4 on Tyreworld (13 actions); most were missing preconditions or effects, not syntax. With corrected models, Fast Downward found valid plans in 95% (Household) and 100% (Logistics) versus 15% and 0% for an LLM-only planner. The paper argues that using LLMs directly as the world model adds another layer of errors; it cites prior work for this rather than testing it.
• articleWhat do later open studies measure about LLM NPC dialogue in games (player ratings, grounding errors, prompt scaffolding), and how small or weak is the evidence for production writing?
  Two open studies were read in full-text excerpt form. Akoury et al. (Findings of EMNLP 2023) had 28 players compare GPT-4 infilled lines against Disco Elysium writers' lines and found designer prose preferred, with 32% of rated generations illogical in context. Figueiredo and Elumeze (arXiv 2510.25820v1) ran a 10-person usability study (no reliable difference after correction) plus a gpt-4o-judged synthetic probe of 60 scripted interactions in which scaffolding helped a quest-giver and hurt suspect NPCs. Both are small, single-game, short-horizon studies; neither measures lore consistency over long play.
• articleWhich multi-agent or multiplayer generative world models can I actually download or run, and what do the repositories and model cards say about their licences?
  Of eight multi-agent world-model papers inspected, Solaris (two-player Minecraft) ships code, evaluation data, training data and weights under Apache-2.0 per its repository and card. Gamma-World ships Apache-2.0 code and .safetensors checkpoints, but its model card states no licence. MultiWorld states code Apache-2.0 and data/weights CC BY-NC 4.0. MIRA releases code and a gated CC BY-NC-SA dataset but advertises no weights in the text read. MetaWorld releases only single-person code. No release was found for MASS, WorldCast or ShareVerse.
• articleWhat did Park et al. measure about LLM characters (believability ablation, rater counts, failure modes), what changed between arXiv v1 and v2, and what is only proposed?
  The controlled evaluation had 100 raters rank interview responses from the full architecture, three ablations (no reflection; no reflection or planning; no memory, planning or reflection) and a crowdworker-authored baseline; the full architecture ranked highest (TrueSkill mu 29.89) and the memoryless ablation lowest. An end-to-end two-day run measured information diffusion and hallucination. Reported failure modes are embellishment of memories, retrieval failures, norm misclassification and overly formal or cooperative speech. A number differs between v1 (12 agents, 48%) and v2 (13 agents, 52%). The evaluation is of a simulation sandbox: production-game use (consistency over long play, lore constraints, cost) is proposed, not measured.
• articleHow does Dramatis compute suspense for a story plan, how well did it match readers, and what part of it can an agent reuse as a stakes/suspense check?
  Dramatis treats suspense as how hard it is, at each point in the telling, for the audience to imagine a plan that lets the protagonist escape an expected bad outcome. It needs a symbolic story encoding, scripts for typical negative outcomes and a memory model that decides which actions seem likely. On three paired film-derived scenes (32 readers), its curve areas ordered each original above its lowered-suspense alternate, matching reader preferences, but one pair (Rear Window) differed only slightly, and ablations did not consistently match readers. The reusable idea for an outline is the escape-plan test: name the expected bad outcome, then ask how easy the most likely escape plan is to imagine and how it is blocked.
• articleHow do the primary papers evaluate latent actions, and which of those checks could an agent re-use on its own model?
  Every paper uses a different test, so numbers cannot be compared across papers. Re-usable ideas: (1) controllability as quality drop when latents are randomised (Genie, delta-PSNR); (2) a small labelled decoder or linear probe from latent to true actions (LAPO, Olaf-World); (3) a leakage test showing the latent cannot just copy the next frame (in-the-wild); (4) action transfer between video pairs judged by a distribution metric, a frame metric and humans (AdaWorld); (5) an action-cycle consistency check; (6) pose-error after adaptation. Most real-video transfer judging is qualitative or small human panels.
• articleWhat do the learned latent actions actually correspond to, and how can I find out what each one does?
  What a latent corresponds to is an empirical question per dataset. LAPO found, on synthetic Procgen data with known actions, clusters that match the true actions in many games and fragments in games with partial observability. Genie authors report consistent latent meanings (for example down, up, left on robot video) by inspection of generated frames. On diverse in-the-wild video, a later study reports actions that are camera-relative and localised in space, not tied to one body. None of these gives a naming for latent codes in advance; you find out by trying each code.
• articleWhat numeric frame-time, refresh-rate and comfort budgets do Valve, Meta and Apple actually state, and what evidence stands behind them?
  For frame-time and comfort budgets, Valve's GDC 2015 talk (90 Hz = 11.11 ms, ~2 ms panel persistence, 2 ms running start from measured 1.5-2.0 ms bubbles, 24.22 ms prediction) and Meta's Quest page (Oct 2024: interactive apps at least 72 FPS, media apps may aim for 60; Rift S 80, Rift 90; ranges 'may vary') state numbers, while Apple's visionOS HIG comfort guidance read is qualitative.
• articleWhat do XAG 107 and the Game Accessibility Guidelines require for input remapping?
  XAG 107 asks games to offer in-game remapping of all controls even where the platform remaps, including analog/digital controls and per-stick axis inversion, sensitivity adjustable by at least 50% of default, full operability with digital input, and attention to long holds (2-3 s or more) and simultaneous presses; the Game Accessibility Guidelines list remappable controls as a Basic guideline and say system-level remapping is a safety net, not a substitute.
• articleWhat do Epic's Derived Data Cache docs say about how the DDC works, shared DDC configuration and diagnosing misses (UE 4.27 vs 5.3)?
  The UE 5.3 DDC documentation describes a fastest-first cache hierarchy with asynchronous fill, disposable content regenerable from .uasset files, shared-DDC setup via a config path or UE-SharedDataCachePath (disabled with -ddc=noshared), and diagnosis through LogDerivedDataCache; apart from a KeyLength node hashing long keys, it does not say how keys are built or what invalidates entries.
• articleHow do I get an Unreal Sequencer cut to and from an editor, and what does the exchange preserve or lose?
  Unreal ships a native Final Cut Pro 7 XML exporter for Sequencer (UE 5.8 API: frame rate, resolution and handle frames as parameters); for OTIO, the community plugin round-trips video tracks, gaps, markers, nesting and linear speed but not audio, transitions, effects, non-linear speed, CDL or image-sequence references, so audio and grading must travel separately.
• articleHow did Netflix engineers make branching video play seamlessly: what is the segment/choice model, what gets preloaded, and how is branch state signalled?
  Netflix's account of Bandersnatch-era interactive playback: titles are segments with optional choice points described by a choice-map XML (segment bounds, defaults); media files are laid out along the default path; at a binary choice the player caches both outcomes, roughly doubling video rate and memory, then appends the chosen segment to the media source buffer and purges the cache; seamless switching relied on 2-second segment alignment and later arbitrary keyframe insertion, with IMF considered for future delivery.
• articleWhich fields in the FrameTimelineEvent trace data does Perfetto require and read, so I can tell whether a captured trace is usable?
  Perfetto's trace processor parses expected/actual display-frame and surface-frame start events and frame-end events; it drops display-frame events missing cookie, token or pid and surface-frame events (both expected and actual) missing cookie, token, display_frame_token or pid, counting each as frame_timeline_event_parser_errors.
• articleWhat do I put in a Perfetto trace config to capture Android FrameTimeline jank data?
  To capture FrameTimeline jank data, a Perfetto config needs the data source android.surfaceflinger.frametimeline (no further fields shown) plus at least one buffer, on Android 12 or later.
• articleWhat does the wildwinter Ink-Localiser (LocaliserLib) do to ink dialogue to make it localisable, what are its ID and export formats, how does a re-run behave, and under what licence text?
  Ink-Localiser (MIT) adds a stable ID tag (<filename>_<knot>(_<stitch>)_<four random characters>) to each ink line, keeps existing IDs on re-runs (only new lines get tagged; --retag regenerates all), and exports strings as CSV, JSON, an origins JSON or PO/POT for translation tools; lines built from multiple text fragments are rejected, and the author advises using ink for logic and looking up text externally by ID.
• articleHow does YarnSpinner-Unreal consume localised strings and does it update them when the dialogue changes?
  The YarnSpinner Unreal plugin imports per-language localisation CSVs (listed in the .yarnproject) as UDataTable assets and reimports a CSV only when both its timestamp and MD5 hash have changed; it does not generate the CSVs or flag stale translations, and handling of renamed/deleted/updated projects' localisation assets is left as TODOs.
• articleHow does YarnSpinner-Godot store translated strings and what happens to them when I edit the source dialogue?
  YarnSpinner-Godot updates each language CSV on 'Update Localizations': removed line IDs are deleted, new IDs are added with empty text, and lines whose base text changed (lock hash differs) get their translation prefixed as needing update, after which the lock is refreshed. The code's marker is '(NEEDS UPDATE) ' whereas the usage doc says '(NEEDS UPDATED)', so a search for the documented marker would miss stale rows.
• articleWhat does Unity's manual actually say about Optimized Frame Pacing on Android, and is there a documented way to measure it?
  Unity 6.0's manual describes Optimized Frame Pacing on Android in one Player settings row (evenly distributing frames for less frame-rate variance) with no measurement method or implementation detail on that page; frame pacing has to be verified with external tools such as Perfetto FrameTimeline.
• articleHow do Unity Addressables decide which bundles players re-download after a content update, and what role do the content state file and catalog hash play?
  Per Addressables 2.3 docs, a full rebuild with a new catalog forces players to re-download all remote bundles, whereas a content update build driven by the original release's addressables_content_state.bin keeps unchanged bundles under their original names so only changed bundles download; code changes require a new player build because bundles carry no code; and a catalog hash file beside the catalog (same name, different extension) limits catalog downloads to real changes.
• articleWhere does Yarn Spinner (v3, Unity) keep dialogue variables, how does its built-in save/load work, and does the save carry any version marker?
  Yarn Spinner for Unity keeps variables in a replaceable VariableStorage (in-memory by default, so unsaved state is lost on scene unload or quit); the built-in save writes an unversioned JSON of six parallel key/value arrays to persistentDataPath, and loading replaces existing variables because SetAllVariables clears by default. The docs' sample Contains returns values == null, the inverse of the documented meaning, so copying it into a custom storage would break existence checks.
• articleDo the Twee 3 specification and the Twine 2 story-format specification define anything about game state, saves, or save compatibility?
  The Twine specifications define story notation, StoryData (IFID, story-format name and version) and story-format packaging with semantic versioning, but nothing about runtime state or saves; save compatibility is left to each story format, and the only cross-format signal is the format name and version recorded in StoryData.
• articleHow do Godot AnimationPlayer and AnimationTree divide playback, what are their time units, and what is OpenToonz's licence status?
  Corrected 2026-10-23 - see C3-B642-07; the original statements below are kept for history and are superseded where the correction says so. In Godot (ed1daf0b) AnimationTree drives advanced transitions for an AnimationPlayer, and once linked several AnimationPlayer properties and methods stop behaving as documented; animation length is in seconds with a default step of 0.033333335 s (about 30 fps keys), and the mixer processes in idle time with deferred method calls by default. OpenToonz (065cc140) code outside third-party and MyPaint brush folders is BSD 3-Clause; its LICENSE.txt contains the full BSD text after third-party notes.
• articleHow do Unity text-serialised scenes/prefabs support merging, and what does the UnityYAMLMerge tool document about clean versus unclean merges?
  Corrected 2026-10-23 - see C3-B642-09; the original statements below are kept for history and are superseded where the correction says so. Unity scenes are a custom YAML subset with one document per object (--- !u!<class> &<fileID>) and in-file {fileID: n} references, floats optionally in IEEE 754 hex for lossless values; UnityYAMLMerge merges scenes and prefabs semantically, with Smart Merge Off/Premerge/Ask, a mergespecfile.txt fallback and documented Git/Mercurial/P4V/SVN configuration. Unmergeable cases and .meta/GUID handling are not described on these pages.

More