Meta tags:
description=ReCA: an inference-time framework that solves multi-shot video extrapolation (MSVE) by allocating context hierarchically across Plan, Allocate, and Refresh operators, achieving 8–16% gains over the strongest baseline.;

Headings (most frequently used words):

shot, reca, multi, video, context, extrapolation, allocation, and, on, long, recursive, is, problem, msve, preview, every, parallel, via, failure, not, length, it, decomposes, into, bounded, jobs, plan, for, the, 18, heavenly, titans, sequence, continuations, across, landmarks, action, food, dialogue, framework, comparison, anchor, leads, backbone, bench, overall, normalised, score, human, raters, rank, first, likert, criterion, if, you, find, useful, please, cite, our, work, segment, execution, root, to, planning, local, global, level, rendering, preparing,

Text of the page (most frequently used words):
the (60), and (49), shot (38), reca (27), video (23), #context (23), msve (20), state (19), long (16), multi (14), extrapolation (11), allocation (11), into (11), that (10), each (10), anchor (10), #recursive (9), per (9), across (9), prompts (8), backbone (8), call (8), leaf (8), generation (7), identity (7), frozen (7), generator (7), frame (7), segment (7), vgot (6), mora (6), movieagent (6), visual (6), narrative (6), over (6), wan (6), method (6), short (6), bench (6), vistory (6), scene (6), its (6), while (6), task (6), observed (6), consistency (5), every (5), with (5), for (5), time (5), planning (5), from (5), object (5), structured (5), demo (4), via (4), liu (4), wang (4), arxiv (4), baseline (4), cross (4), first (4), 978 (4), quantitative (4), source (4), block (4), duration (4), only (4), action (4), continuation (4), sequence (4), single (4), render (4), level (4), parallel (4), bounded (4), boundary (4), shots (4), execution (4), cinematic (4), one (4), structure (4), cinematically (4), interactive (3), under (3), 2605 (3), 26525 (3), character (3), where (3), below (3), raters (3), user (3), study (3), movie (3), methods (3), clip (3), ours (3), i2v (3), extension (3), avg (3), moviebench (3), self (3), storymem (3), vbench (3), evaluation (3), proprietary (3), check (3), happyhorse (3), open (3), strongest (3), score (3), same (3), budget (3), prompt (3), allocates (3), temporal (3), are (3), generated (3), framework (3), preview (3), scale (3), plan (3), timeline (3), once (3), independent (3), then (3), segments (3), previous (3), tail (3), memory (3), budgets (3), global (3), intent (3), inside (3), frames (3), can (3), which (3), minute (3), challenge (3), but (3), than (3), preserving (3), models (3), three (3), failure (3), university (3), github (2), internal (2), review (2), akide (2), xing (2), jinbo (2), mao (2), chaojie (2), zhang (2), zeyu (2), yefei (2), weijie (2), zihan (2), haffari (2), gholamreza (2), zhuang (2), bohan (2), preprint (2), 2026 (2), our (2), compare (2), six (2), human (2), criteria (2), background (2), criterion (2), gaps (2), next (2), mean (2), four (2), likert (2), metrics (2), 913 (2), 990 (2), 798 (2), 974 (2), 572 (2), generality (2), generalisation (2), 062 (2), 992 (2), 263 (2), 988 (2), 454 (2), all (2), purpose (2), built (2), average (2), normalised (2), local (2), paper (2), results (2), compared (2), handoff (2), causality (2), comparison (2), model (2), compact (2), range (2), through (2), several (2), preparing (2), expands (2), final (2), system (2), planner (2), whole (2), stay (2), allocate (2), schedule (2), keeps (2), story (2), calls (2), subproblems (2), back (2), plans (2), window (2), decomposes (2), existing (2), paradigms (2), address (2), fragments (2), preserves (2), lacks (2), storytelling (2), imposes (2), yet (2), remains (2), free (2), invent (2), states (2), rather (2), continue (2), generators (2), extends (2), advancing (2), finite (2), usable (2), bottlenecks (2), unsupported (2), lossy (2), chaining (2), propose (2), inference (2), hierarchically (2), this (2), improves (2), not (2), length (2), problem (2), abstract (2), code (2), bibtex (2), tongyi (2), lab (2), alibaba (2), group (2), article, liu2026reca, title, author, journal, year, eprint, archiveprefix, primaryclass, url, https, org, abs, copy, you, find, useful, please, cite, work, citation, grouped, bars, rating, appeal, script, faithfulness, physical, law, coherence, wins, order, widest, best, stays, whiskers, report, cell, standard, error, higher, better, ratings, automated, videos, min, scoring, random, subsets, hidden, largest, sit, dimensions, saturated, cannot, distinguish, rank, 737, 517, 383, 866, 985, 636, 687, 962, 284, 610, 980, 976, 600, 273, 165, 331, 975, 153, 470, 370, 710, 485, 102, 214, 158, 539, 961, 948, hosted, vendors, 683, 819, 378, 755, 891, 608, 513, 232, 753, 945, 878, 545, 187, 726, 303, 666, 885, 520, 113, 144, 631, 871, 094, 327, 127, 512, 943, 853, primary, reproducibility, target, real, factor, lowest, among, gain, metric, 749, 942, 324, 936, 984, 691, 738, 467, 286, 602, 276, 855, 972, 569, 182, 959, 274, 965, 492, 123, 881, 142, 587, 952, 951, within, baselines, share, package, differs, how, shown, default, available, numbers, reproduced, table, leads, overall, continuity, sequences, performance, case, starts, premise, asks, carry, camera, changes, transitions, beats, continuations, landmarks, food, dialogue, featured, cases, select, beat, inspect, role, selected, duel, cloud, combat, titan, transformations, standoff, keeping, characters, weapons, readable, heavenly, titans, contexts, assigned, job, walk, serially, rendering, receives, slice, location, references, most, recent, entries, packed, message, never, sees, token, starting, multiple, anchors, builds, semantic, goals, durations, conditions, dependency, links, instead, sending, root, parallelism, whose, together, serial, executable, divided, budgeted, unit, chooses, goal, condition, preserved, separates, decides, units, need, run, terminology, phases, recursion, induces, produce, runs, leaves, chains, conditioned, ones, sequentially, extracts, verifies, refreshes, external, feeds, adaptive, update, logic, outside, tree, writes, pick, last, left, off, jobs, capability, chapters, bridges, these, two, regimes, carries, forward, fits, overview, definition, narrower, nominal, panels, diagnose, detailed, introduce, constraints, longer, reduce, salience, needed, current, become, aging, additional, helps, raises, density, dilution, overload, failures, recursively, invokes, nodes, propagates, updates, evaluate, setting, further, grounded, protocol, normalized, competing, controller, identify, coupled, planners, specify, details, full, screenplays, dilute, relevant, when, carrying, complete, turns, decay, reveals, merely, limitation, central, generative, define, release, progress, repository, will, populated, com, ali, vilab, monash, zhejiang, queensland, corresponding, authors, intern, coherent, event, allocating, operators, refresh, try, showcase,

Text of the page (random words):
reca multi shot long video extrapolation via recursive context allocation reca abstract msve method timeline showcase comparison results user study bibtex preparing demo try interactive demo recursive context allocation preprint 2026 reca multi shot long video extrapolation via recursive context allocation reca is an inference time framework that expands an observed visual anchor into a coherent cinematically structured multi shot continuation preserving identity scene object and event causality state by allocating context hierarchically across plan allocate and refresh operators akide liu 1 2 3 jinbo xing 2 chaojie mao 2 ye li 3 zeyu zhang 3 yefei he 3 weijie wang 3 zihan wang 4 yu liu 2 gholamreza haffari 1 bohan zhuang 3 1 monash university 2 tongyi lab alibaba group 3 zhejiang university 4 university of queensland corresponding authors intern at tongyi lab alibaba group paper code interactive demo bibtex code release in progress of internal review repository will be populated at github com ali vilab reca abstract long video failure is not a context length problem it is a context allocation problem minute scale cinematic video generation is a central challenge for generative video models existing paradigms address only fragments of this challenge single shot extrapolation preserves an anchor but lacks cinematic structure while multi shot storytelling imposes structure yet remains free to invent its visual states rather than continue an observed one we define multi shot video extrapolation msve a task that extends an observed frame or clip into a sequence of cinematically structured shots while preserving anchor state and advancing narrative intent under the finite per call generation budget of short video models we identify three coupled bottlenecks 1 global planners over specify unsupported details from full screenplays 2 shot level prompts dilute task relevant state when carrying the complete story 3 temporal chaining turns generated frames into a lossy memory in which identity scene object and action state decay msve reveals that long video failure is not merely a limitation of context length but a failure of context allocation we propose recursive context allocation reca an inference time framework that allocates context hierarchically across planning and generation reca recursively decomposes msve into context bounded subproblems invokes frozen generators at leaf nodes and propagates structured state updates across time to evaluate this setting we further propose msve bench and nb q a source grounded protocol with prompts purpose built for 3 5 minute long video generation compared to previous methods reca improves average normalized score by 8 16 over the strongest competing controller and improves multi shot consistency metrics by 28 43 context allocation failures in msve the usable context of a frozen video generator is narrower than its nominal window the three panels diagnose the three bottlenecks a global planning overload over detailed plans introduce unsupported constraints b shot level dilution longer per shot prompts reduce the salience of state needed by the current call and c temporal chaining generated frames become a lossy aging memory of identity scene object and action state additional context helps only while it raises the density of usable state task definition multi shot video extrapolation msve msve extends an observed frame or clip into a sequence of cinematically structured shots while preserving anchor state and advancing narrative intent under the finite per call generation budget of short video models msve task overview existing paradigms address only fragments of the minute scale generation challenge single shot extrapolation preserves an anchor but lacks cinematic structure while multi shot storytelling imposes structure yet remains free to invent its visual states rather than continue an observed one msve bridges these two regimes it carries the observed anchor s identity scene and object state forward across a cinematically structured continuation that fits inside the per call duration and prompt budgets of short video generators capability chapters reca decomposes multi shot video extrapolation into context bounded shot jobs the method keeps task state outside the frozen video generator plans the long continuation as a tree of shots allocates a per shot context window to each leaf and writes the tail frame back into state so the next leaf can pick up where the last one left off recursive context allocation reca four phases the recursion induces 1 recursive planning of the long video task into context bounded subproblems 2 leaf execution where each leaf calls the frozen generator to produce one segment 3 leaf execution logic that runs independent leaves in parallel and chains boundary conditioned ones sequentially and 4 adaptive state update that extracts verifies and refreshes external state then feeds back to the planner terminology shot segment and parallel execution reca separates the movie timeline from the short video calls that render it then decides which units need a boundary handoff and which can run at the same time shot a cinematic unit in the final long video the system chooses its goal duration boundary condition and preserved state segment an executable leaf call to the frozen short video generator a long shot can be divided into several budgeted segments parallel execution parallelism is at the shot level shots whose anchor frames are independent render together through the frozen generator segments inside a single shot stay serial each segment s first frame is the previous segment s tail 01 plan root to shot recursive planning starting from multiple visual anchors and narrative intent reca builds a shot schedule with semantic goals durations boundary conditions and dependency links the schedule keeps each generator call inside prompt and duration budgets instead of sending the whole story at once 02 allocate local global context allocation each shot receives a compact context slice character location identity references the anchor frame and the most recent state memory entries packed into a per shot system message the planner never sees the whole movie at once so per call token budgets stay bounded 03 render shot level parallel rendering once contexts are assigned each shot is an independent leaf job render the anchor frame first then walk its segments serially each segment s first frame is the previous segment s tail shot preview timeline shot preview plan for the 18 shot heavenly titans sequence the sequence expands a single duel into cloud combat titan scale transformations and a final standoff while keeping the same characters weapons and action state readable selected shot preparing shot preview select a beat to inspect its role in the long range continuation featured reca cases multi shot continuations across landmarks action food and dialogue each case starts from a compact visual premise and asks the model to carry long range state through several camera changes scene transitions and narrative beats model performance compare framework comparison on multi anchor video extrapolation reca is compared with vgot mora and movieagent on identity consistency scene handoff object state continuity and narrative causality across long generated sequences quantitative results reca leads every backbone on msve bench and overall normalised score mean se across 20 msve prompts within each backbone block reca and all baselines share the same frozen generator per call duration budget and long video prompt package reca differs only in how it allocates planning shot local and temporal context the wan 2 7 block is shown by default the open source wan 2 2 block and the happyhorse 1 0 generality check are available below numbers reproduced from the paper s table 1 quantitative evaluation on wan 2 7 proprietary backbone 20 msve prompts method vbench storymem vistory self vistory cross moviebench msve bench avg i2v extension 0 951 0 952 0 587 0 142 19 881 0 123 0 492 vgot 0 978 0 965 0 798 0 274 21 959 0 182 0 569 mora 0 972 0 988 0 855 0 263 26 062 0 276 0 602 movieagent 0 974 0 978 0 913 0 286 25 467 0 738 0 691 reca ours 0 984 0 992 0 936 0 324 31 454 0 942 0 749 8 16 average normalised score over the strongest baseline on every backbone block 28 43 msve bench gain over the strongest baseline on the metric purpose built for multi shot extrapolation 4 74 real time factor the lowest among all methods 18 22 below every baseline wan 2 2 backbone open source generalisation 20 msve prompts quantitative evaluation on wan 2 2 open source backbone our primary reproducibility target method vbench storymem vistory self vistory cross moviebench msve bench avg i2v extension 0 853 0 943 0 512 0 127 19 327 0 094 0 454 vgot 0 871 0 988 0 631 0 263 25 144 0 113 0 520 mora 0 885 0 990 0 666 0 303 23 726 0 187 0 545 movieagent 0 878 0 945 0 753 0 232 26 513 0 572 0 608 reca ours 0 891 0 992 0 755 0 378 26 062 0 819 0 683 happyhorse 1 0 backbone proprietary generalisation check 20 msve prompts quantitative evaluation on happyhorse 1 0 proprietary backbone generality check across hosted vendors method vbench storymem vistory self vistory cross moviebench msve bench avg i2v extension 0 948 0 961 0 539 0 158 20 214 0 102 0 485 vgot 0 978 0 978 0 710 0 370 24 470 0 153 0 572 mora 0 974 0 975 0 798 0 331 25 165 0 273 0 600 movieagent 0 976 0 980 0 610 0 284 27 962 0 687 0 636 reca ours 0 985 0 990 0 866 0 383 28 517 0 913 0 737 user study human raters rank reca first on every likert criterion four methods vgot mora movieagent reca over the 20 msve prompts on wan 2 7 80 long videos at 3 5 min each six 0 5 likert criteria raters scoring random subsets with method identity hidden the largest gaps sit on the cross shot dimensions that saturated short clip metrics cannot distinguish user study ratings for automated movie generation grouped bars compare vgot mora movieagent and reca on six human rating criteria visual appeal script faithfulness character consistency background consistency physical law and narrative coherence reca wins every criterion 4 27 3 99 4 25 3 62 3 74 3 76 in that order with the widest gaps on character and background consistency where the next best baseline stays below 2 1 whiskers report per cell standard error of the mean across raters higher is better citation if you find reca useful please cite our work copy article liu2026reca title reca multi shot long video extrapolation via recursive context allocation author liu akide and xing jinbo and mao chaojie and li ye and zhang zeyu and he yefei and wang weijie and wang zihan and liu yu and haffari gholamreza and zhuang bohan journal arxiv preprint arxiv 2605 26525 year 2026 eprint 2605 26525 archiveprefix arxiv primaryclass cs cv url https arxiv org abs 2605 26525 reca multi shot long video extrapolation via recursive context allocation github under internal review interactive demo