long-horizon-llm-agents-environments-where-erroneous-action-changes-world CLM-0431For long-horizon LLM agents in environments where an erroneous action changes world state, restoring an earlier checkpoint and carrying forward a structured reflection memory that excludes branch-local state claims beats either intervention alone, because reflection alone cannot undo environmental damage and rollback alone re-explores the same failure.
Capability: Completing tasks that take many steps or many hours
Observed on
Training-free test-time control on ALFWorld, ScienceWorld and GAIA with Qwen3-14B and DeepSeek-V3, where environment and workspace state can be checkpointed and restored; same LLM .
Sources
- supportsRollback the World, Keep the Reflection: Rollback-Induced Reflection for Long-Horizon LLM AgentsValidated by: Success rate on ALFWorld, ScienceWorld (plus dense reward) and GAIA against ReAct, Reflexion, LATS, Self-Refine and GA-Rollback under matched budgets: average SR 47.00 vs 43.91 best baseline (Qwen3-14B) and 69.43 vs 62.86 (DeepSeek-V3); ablations give 76.75 SR vs 70.48 reflection-only and 69.37 rollback-only on ScienceWorld, with 0.92 rollbacks per task vs 2.9 for GA-Rollback.. Measured on three benchmarks with two backbones and component ablations that isolate rollback from reflection; [truncated]
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- In LLM multi-agent systems that reflect after failure, restricting reflection to the single agent responsible for the first incorrect step — located by a trained step-level correctness scorer and paired with a counterfactually generated corrected step — raises task success more than having every agent reflect on the trajectory, because reflections written by agents that behaved correctly enter their memory as misleading guidance for later tasks.Getting more from several models than from one · reviewed by AI
- For LLM multi-agent systems failing on long-horizon web tasks, re-running the whole task from scratch — plain, with self-reflection, or with a critic agent — repairs few failures (at most 6.90% within three attempts, with feedback variants below plain rerun), while reconstructing the recorded execution prefix and intervening once at a symptom-flagged trace node repairs 20.15% with a single attempt.Getting more from several models than from one · reviewed by AI
- For LLM agents acting through stateful tools over many turns, unsafe policy-violating completions accumulate with the interaction budget and are roughly doubled when the adversarial user conditions each next request on the agent's actual prior response, so single-turn or short-prefix safety evaluation understates the failure rate.Using the tools it is given · reviewed by AI
- For single-agent procedural tasks where the relevant state fits a fixed schema known in advance, replacing the appended conversation transcript with a structured execution state that the model rewrites each step — discarding the reasoning trace once the state patch is validated — holds accuracy flat as the horizon grows while history-appending runtimes degrade, and the gain comes from the structure rather than the shorter prompt.Completing tasks that take many steps or many hours · reviewed by AI
- For LLM agents on dependent multi-step tasks, success decays geometrically with the number of steps under a single per-step reliability parameter that rises with model scale but stays below 1, so long horizons eventually collapse even for the strongest deployed systems — and the decay tracks step count, not context length, since bounding the context window steepens the decay rather than easing it.Completing tasks that take many steps or many hours · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: An ablation on the same benchmarks where reflection-only or rollback-only matches or exceeds the combined method at equal interaction and auxiliary-call budget, or where the gain disappears once the memory is allowed to include current-state claims.