llm-web-agents-cluttered-pages-containing-many-concurrent-interface CLM-0427For LLM web agents on cluttered pages containing many concurrent interface components, success falls sharply as component count grows, and having the agent first emit a localized sub-goal plus a task-relevant sub-observation before choosing an action recovers much of the loss, because self-assessment of the action is more reliable in the reduced context.
Capability: Using the tools it is given
Observed on
Text/accessibility-tree web environments (CompWoB+, WebArena, WorkArena) with an 8B open model and GPT-4o-mini; label-free adaptation via ICL for single-turn static pages or RL for.
Sources
- Validated by: Task success rate against no adaptation, SFT/RL with ground-truth rewards, TTT and TTRL: on CompWoB+ at 8 components 19%->61% (Qwen3-8B) and 46%->88% (GPT-4o-mini); on WebArena 12.1% base, 14.8% RL with ground truth, 12.5% TTT, 21.0% TTED-RL; WorkArena out-of-distribution 18.2%->22.7%. Ablations: removing sub-goal drops to 12.4%, sub-observation to 15.0%. Human agreement with reward labels 0.85 for TTED vs 0.71 TTT, 0.68 TTRL.. Measured on one synthetic benchmark (600 tasks) and a WebArena subset plus WorkArena, mainly with Qwen3-8B; [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
- For LLM agents on long-horizon web penetration testing, holding the backbone model fixed and moving global state out of the context window onto a persistent graph of verified facts and fact-conditioned intent edges raises task success most on the hardest tasks, where success depends on facts obtained dozens of steps earlier (88.2% overall and 75.0% hard, vs 44.1% and 25.0% for the strongest shared-plain-text-memory baseline).Completing tasks that take many steps or many hours · reviewed by AI
- 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 agents that score their own actions during exploration in text-based games and then reuse that experience, the accuracy of those self-scores against the environment verifier does not predict later gains on held-out stricter configurations: block-level correlations between judgment agreement or calibration and the next checkpoint's score change sit near zero, so recognizing a good action is not the bottleneck — converting it into a transferable policy is.Improving its own scaffolding from failures · 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
- In multi-agent LLM workflows with a mixed-size model zoo, choosing which model instantiates each agent step-by-step from an online estimate of task-progress gain plus remaining time and cost budget keeps task quality near the best single-model policy at lower energy than one-shot query-level routing or stall-triggered escalation.Getting more from several models than from one · 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: A replication where sub-goal/sub-observation decomposition gives no gain over undecomposed test-time training at matched budget, or where decomposition does not improve agreement between the agent's self-assessed rewards and human judgements.