Search: “program-solve”

Smart search ranks by meaning as well as keywords (one row per story, last 45 days).

Towards a Deterministic Math Solver for Clinical Language Models

Paper shows handing arithmetic to a deterministic Python solver beats direct model calculation at 32B but not reliably at 7B on MedCalc-Bench.

Researchers test a Program-Solve interface where clinical LLMs write case-specific Python executed by a restricted local solver instead of doing arithmetic directly. On MedCalc-Bench Verified (1,100 cases, 55 calculators), Qwen2.5-32B-AWQ scored 90.53% with solver handoff versus 83.47% with direct arithmetic (+7.05 points), while Qwen2.5-7B gained an unreliable +3.29 points with a confidence interval spanning zero. The authors audited the benchmark against clinical guidelines and flagged 16 of 55 calculators for version, use, or coefficient concerns.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

25

Beyond Solver Verdicts: Generative Reward Models for Autoformalization

Researchers introduce Generative Verification (GenV), a generative reward model achieving 0.961 AUROC in detecting unfaithful autoformalization that preserves solver verdicts.

The paper formalizes Verdict-Preserving-Unfaithfulness (VPU), a failure mode in neurosymbolic autoformalization where an incorrect encoding executes successfully and matches the expected solver verdict, and proves verdict-only verification is bounded to chance-level detection. The proposed Generative Verification (GenV) distills an offline Z3-equivalence oracle into a reference-free, continuous reference-equivalence score within the language model's vocabulary space. The oracle-mined verifier (GenV+HN) achieves 0.961 AUROC, generalizes zero-shot across unseen translators and formal styles, and yields an 11.3-point downstream accuracy gain in agentic test-time compute allocation. Mechanistic analysis with decision-projected logit lenses and sparse autoencoders shows the generative readout extracts precise spatial error coordinates without explicit localization training.

Hugging Face daily papers·Sep 9, 20:00 UTC · 6d agoAI research1

35

Root-Cause Attribution Is a Search Problem: Continual Search for Long-Horizon Agent Failures

Continual Search framework iteratively prompts LLM judges to keep searching agent execution logs, boosting long-horizon failure root-cause attribution accuracy.

The paper frames automated root-cause attribution (RCA) for long-horizon AI agent failures as a search problem, since relevant evidence is sparse and distributed across massive execution traces. The authors propose Continual Search, an iterative framework that nudges an LLM judge across successive turns to keep hunting unresolved diagnostic evidence instead of settling on an early plausible diagnosis. They introduce MegaRCA-Mix, a benchmark of 50 human-annotated failure trials on long-horizon, execution-heavy tasks. On MegaRCA-Mix, Continual Search improves GPT-5.5's F1 from 0.349 to 0.498 (over 40% gain), and lower-tier models can surpass higher-tier counterparts when search is effective.

Hugging Face daily papers·Sep 10, 20:00 UTC · 5d agoAI research1

42

Diffs vs. Whole Files: An Empirical Comparison of Iterative Edit-Based and Direct Generation for Flutter/Dart Code Models

Empirical study finds direct whole-file generation beats iterative diff-based editing for Flutter/Dart code models on about 1,790 held-out tasks.

Researchers trained Rainbow-Pony-100M from scratch and fine-tuned Qwen2.5-Coder-0.5B in both direct-generation and diff-based regimes, then evaluated four resulting models on roughly 1,790 Flutter/Dart tasks. Direct generation outperformed diff-based generation on compilation pass rate, bits-per-byte, character-level similarity, and blinded LLM-judge ratings. Diff-based editing is competitive only on short, localized edits in refactoring and error-handling tasks, a property the authors call task locality.

Hugging Face daily papers·Sep 4, 20:00 UTC · 11d agoAI research1

30

Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization

Researchers release OR-Clarify, a benchmark testing whether LLM agents ask clarifying questions before formulating optimization models from incomplete requests.

OR-Clarify evaluates pre-formulation clarification in operations research: each task gives a partial problem description, withholds structured hidden slots, and scores agents via bounded interaction with a simulated user, measuring slot recovery, stopping behavior, silent assumptions, and interaction cost. The authors also propose InterOPT, a two-stage framework that identifies formulation-critical gaps to decide when to ask or stop. In choice-based experiments InterOPT substantially outperforms all baselines in exact slot recovery and remains competitive in the open-ended setting.

Hugging Face daily papers·Sep 3, 20:00 UTC · 12d agoAI research1

35

ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Search

ZGCM-1 is a fully open 7B foundation model with 256K context that stays competitive with frontier models on math reasoning and agentic search.

ZGCM-1 is a fully open 7B dense foundation model trained from scratch using an efficiency-focused recipe: interleaved gated sliding-window and full attention, a stable FP8 Muon optimizer, and MDP-based mid-training with context scaling across 16K, 64K, and 256K. On mathematical reasoning and agentic search suites it remains competitive with much larger frontier models such as Qwen3-235B-A22B and GLM-5.1. The recipe yields a ~4.2x improvement in 16K pre-training time-to-loss, and all weights, checkpoints, training code, data recipes, and W&B logs are open-sourced.

Hugging Face daily papers·Sep 10, 20:00 UTC · 5d agoModel release

50

PuzzleMask: Abusing Plain Prose as a Covert AI Attack Vector

Check Point details PuzzleMask, a plain-prose technique that bypasses LLM gatekeeper policy checks, letting hidden payloads reach target models unreviewed.

Check Point Research describes PuzzleMask, a prompt-crafting technique that hides policy-violating payloads inside plain-English prose wrappers, bypassing quick LLM-based policy checks without emojis, Base64, or invisible formatting. The researchers tested 23 automated prompts against gatekeepers including GPT-4o-mini, GPT-OSS-Safeguard 20b, Claude 3 Haiku, and Llama Guard 3, and all were classified as safe despite policies that flagged the plain versions. When submitted to GPT-5 in thinking-high mode with a Python interpreter, the target model extracted and acted on the payload in over 90% of trials. The technique is not itself a jailbreak but can carry a jailbreak prompt as payload; mitigations include input paraphrasing, hardened gatekeeper policies, and output monitoring.

Check Point Research·updated Sep 11, 05:53 UTC · 5d ago·first Sep 10, 14:32 UTC · 5d agoAI safety & security 2 sources

55

I accidentally turned LLM memory into program analysis

A pwning.systems write-up describes how LLM memory functionality was unexpectedly repurposed into a program analysis technique.

A security research post on pwning.systems describes the author's discovery that LLM memory behavior effectively functioned as program analysis. The write-up is hosted on a security-focused blog and surfaced via a security-tagged link aggregator. Detailed technical content is not included in this feed, limiting verifiable specifics.

Lobsters · security·Aug 29, 03:16 UTC · 18d agoResearch1

30

Can your coding style predict whether your code is vulnerable?

University of Massachusetts Dartmouth researchers present VulStyle, a stylometry-based vulnerability detector that also exposes benchmark reliability problems.

VulStyle combines stylometric features with syntax-tree structure and source tokens, pre-trained on about 4.9 million functions across seven programming languages and fine-tuned on five vulnerability detection datasets. It beat token-only detectors on some benchmarks but its F1 drops sharply on DiverseVul, which the authors link to noisy labels inflating reported performance across popular datasets. The authors argue style-aware detection should be harder to evade but did not test this empirically, and they note that uniform LLM-generated code may strip away the individual developer style the model depends on.

Help Net Security·Aug 24, 04:21 UTC · 23d agoResearch1

48

Retrofitting Code Using LLMs to Support Exceptional Behavior

EXCODER combines static/dynamic analysis with LLMs to retrofit exception-handling code, achieving 85.92% pass@1 with Qwen 2.5 Coder 32B on Java benchmarks.

The paper introduces the task of retrofitting existing code with Exception Related Code (throw statements, guarding conditions, try/catch blocks) so that given Exceptional Behavior Tests pass. EXCODER performs context engineering by integrating static and dynamic program analysis output with LLMs; it was evaluated on a benchmark built from 304 methods across 75 GitHub Java projects. Combined with Qwen 2.5 Coder 32B, EXCODER achieves pass@1, 5, and 10 rates of 85.92%, 86.18%, and 86.51%, roughly 13 percentage points over baseline, and manual inspection reveals remaining limitations.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:18 UTC · 6d agoAI research1

25

MasterControl Seventeen Every Time

Governed enterprise analytics study shows deterministic policy execution matched 110/110 answer-and-evidence contracts while runtime agent planning matched none.

The paper studies a governed approach where a language model interprets the question while deterministic policy selects and runs a pre-approved analytical program returning results and evidence. Across 440 runs, three 8B models generated SQL and selected tools at runtime, while Qwen3-8B only interpreted intent and policy executed the approved program. None of 330 runtime-planning episodes satisfied the full answer-and-evidence contract, whereas the policy-executed analyzer matched 110 of 110. The authors note this is configuration-specific and expressiveness is preserved via relational operations, aggregation, comparison, windows, ranking, and similarity with replayable results.

Hugging Face daily papers·Sep 1, 20:00 UTC · 14d agoAI research

24

Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye

METR analysis finds AI accelerating cyber vulnerability discovery, while SPADE self-play environment generation improves Qwen3 reasoning benchmark scores at 30B scale.

Import AI 470 discusses a METR research note reporting differential acceleration from AI: major acceleration in reported cyber vulnerabilities (cURL, OpenSSL, Firefox, Microsoft, NVD, OSV), minor acceleration in mathematics, and no measurable acceleration in AI-research optimization benchmarks. It also covers SPADE, a self-play framework from a multi-university team (University of Washington, Stanford, MIT, CMU, and others) that co-evolves executable training environments and agent capability using Environment Designer and Reasoning Agent roles with hint-based regret rewards. Trained on Qwen3-4B-Instruct-2507, Qwen3-8B, and Qwen3-30B-A3B-Instruct-2507 via GRPO (400 rollouts of 25 environments), SPADE lifted the 30B-A3B game-environment suite average to 58.3, +8.1 over base, and improved tool-use results across backbones. The issue also references Hawkeye for building better GPU kernels.

Import AI·Aug 24, 13:12 UTC · 22d agoAI research

42

Google Open-Sources Mantis: A Modular Skills Toolkit That Lets Coding Agents Find, Reproduce and Patch Vulnerabilities

Google open-sourced Mantis, an Apache-2.0 modular skills toolkit that lets AI coding agents find, reproduce, and patch vulnerabilities with sandboxed verification.

Google released Mantis on GitHub under Apache 2.0 as a stack-agnostic set of slash-command skills that chain through the full vulnerability lifecycle: mining version history, building threat models, filtering findings, reproducing bugs in gVisor or network-disabled VMs, assembling exploit chains, patching, and scoring residual risk from 1 to 10. It runs with Gemini CLI, Antigravity CLI, the Google ADK, or comparable agent frameworks, and a supervisor skill (/mantis-meta-agent) can drive the whole loop. Google says the design targets the sub-7 percent true-positive rate of naive AI code scanning, and that its hierarchical summary tree cuts token overhead by over 85 percent. The toolkit is deployable for local and internal evaluation but not yet recommended for production.

MarkTechPost·Sep 9, 22:58 UTC · 6d agoAI tools & infra

55

Efficient Branch-and-Bound Testing and Verification of zkVMs

ZEBRA verifies zkVM constraint systems via branch-and-bound cardinality counting, finding 11 zero-day bugs across five real-world zkVMs and running 51.5x faster than SMT verification.

ZEBRA reduces zkVM correctness to a solution-set cardinality problem requiring that each constraint system admit exactly one valid execution trace, eliminating redundancies like null-row padding and non-deterministic permutations before counting. It lifts analysis from finite-field witnesses to an integer interval lattice, exploiting that constraints across 5 real-world zkVMs use only 14.0% of theoretical connectivity capacity on average, enabling tight interval propagation. A parallel branch-and-bound search produces concrete counterexamples or certifies absence of violations within a bounded region. ZEBRA discovers 11 zero-day bugs (6 independently confirmed, 3 fixed), is 51.5x faster than SMT-based verification, and verifies 16.5 percentage points more instances.

arXiv cs.CR·Sep 14, 04:35 UTC · 2d agoResearch

45

AI models' written reasoning steps correspond to distinct internal patterns, a new study finds

KAIST and Naver AI Lab researchers show LLM reasoning steps like extraction and computation map to distinct activation patterns, strongest in middle layers.

Researchers at KAIST and Naver AI Lab defined eight recurring reasoning operations, including extraction, decomposition, formula recall, deduction, and computation, and showed they correspond to separable activation patterns in Qwen2.5-7B, Qwen3-8B, and Gemma4-31B on math tasks, with GPT-5 labeling solution segments. The separation peaks in middle layers, holds even when a computation step produces a wrong answer, and goes beyond surface-level token choice. Findings replicated on Llama-3-8B, and classifiers trained on Qwen3-8B transferred to GPQA-Diamond and MATH-500. The authors note that using internal states for error detection or mid-generation steering remains future work.

The Decoder·Sep 12, 13:39 UTC · 3d agoAI research2

42

An Empirical Analysis of CodeQL False Positives and Query Refinements for Java Vulnerabilities

Study of 167 Java CVE instances finds CodeQL false positives follow recurring patterns; query refinements remove 81.8% of reviewed ones.

Researchers ran CodeQL's Java security query suite on 167 CVE instances from 110 projects, manually reviewing 500 sampled false-positive paths and building a five-category taxonomy led by Missed Path Constraint or Sanitization (36.6%), Benign Execution Context (29.4%), and Missing Trust Boundary Modeling (27.6%). Guided by the taxonomy, query-level refinements removed 81.8% of reviewed false positives and 15.8% of reported paths across the selected queries while retaining 7 of 8 true positives. To address generalization, agentic coding tools given the refinement patterns as templates adapted them to new projects successfully in 56% and 62% of tasks, versus 28% without guidance.

arXiv cs.CR·Sep 3, 22:48 UTC · 12d agoResearch1

34

Studying Without a Syllabus: Task-Agnostic Environment Preprocessing

Paper formalizes task-agnostic environment preprocessing, where agents study unfamiliar environments under a budget to build reusable artifacts for a frozen solver.

The paper formalizes task-agnostic environment preprocessing, where a studying system explores an environment under a budget and produces artifacts like indices, scripts, or procedural guidance for a frozen solver, without task examples or evaluation feedback. The authors compare unaided and archive-equipped meta-agents against fixed synthetic-practice and corpus-processing methods across six heterogeneous benchmarks. A meta-agent variant achieves the highest Avg@3 reward on five benchmarks, while fixed corpus processing remains best on the largest corpus benchmark. Studied artifacts reduce the test-time sampling needed to reach a given score, shifting computation from repeated test-time attempts to a pre-task study phase.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

20

Adversarial Testing of Automated Program Repair Agents for Security Vulnerabilities

SWEADV benchmark shows adversarial issue descriptions make LLM program-repair agents write insecure fixes in 51.7% of cases, evading most detection tools.

Researchers built SWEADV, a benchmark of 750 adversarial issue descriptions derived from 150 SWE-bench Verified repair tasks, covering command execution, deserialization, path traversal, denial of service, and weak hashing attack types. Tested on mini_swe agents backed by GPT-5-Mini, MiniMax-M2.5, and DeepSeek-R, adversarial descriptions induced malicious behavior with successful repair in 51.7% of cases. Detection was weak: LLM-as-judge pre-repair screening reached only 62.3% accuracy, and post-repair detection via static analysis and LLM-as-judge achieved just 39.4% and 55.4%.

arXiv cs.CR·Sep 14, 17:53 UTC · 1d agoAI safety & security1

52

When Models Edit Too Much: On the Fidelity of Minimal Code Edits

A 400-task BigCodeBench evaluation shows frontier LLMs widely over-edit code; a preservation instruction cuts excess edits and raises Pass@1 by 2.3 points.

Researchers built an evaluation framework from 400 BigCodeBench problems with injected AST-level corruptions, each with a known minimal patch, to measure over-editing in LLM code repair. Even strong models like GPT-5.5 produce unnecessarily large edits despite high Pass@1. Adding a preservation instruction reduced average excess Levenshtein distance from 0.195 to 0.131, cut added cognitive complexity by 26.6%, and raised Pass@1 by 2.3 points. Reinforcement learning post-training gave the best out-of-domain edit-fidelity trade-off, while supervised fine-tuning overfit to seen corruption patterns.

Hugging Face daily papers·Sep 2, 20:00 UTC · 13d agoAI research1

30

Procedural Graphs: Self-Evolving Execution Structures for LLM Agents

Researchers introduce Procedural Graphs, self-evolving (procedure, relation, procedure) structures guiding LLM agent tool use and planning.

Procedural Graphs organize procedural knowledge into (procedure, relation, procedure) triplets to guide LLM agent actions, addressing drift such as lost objectives, out-of-order tool calls, and repeated unproductive steps. At each decision step the framework localizes the active node and a guidance model translates the surrounding subgraph into step-level situational guidance. An LLM refiner edits graph topology by contrasting failed with successful trajectories, and across datasets, task types and LLMs the approach outperforms memory-based baselines and matches or surpasses hand-designed graphs.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:59 UTC · 7d agoAI research1

25

Procedural Graphs: Self-Evolving Execution Structures for LLM Agents

Procedural Graph framework stores procedural knowledge as triplets and self-evolves via LLM refinement, beating memory-based baselines across datasets, tasks, and LLMs.

The Procedural Graph organizes procedural knowledge into (procedure, relation, procedure) triplets; at each decision step the framework localizes the agent's active node and a guidance model translates the surrounding subgraph into step-level guidance that biases the solver's next action. An LLM refiner contrasts failed with successful trajectories and edits the graph's topology and attributes, retaining rejected edits to discourage repetition. Starting from a minimal skeleton, the loop builds graphs that match or surpass hand-designed ones and can repair flawed expert priors, delivering consistent gains over memory-based baselines across multiple datasets, task types, and LLMs.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI research1

36

ExecCritic: Learn to Test, Test to Improve for Coding Agents

ExecCritic separates test generation from patching for coding agents, lifting SWE-bench Verified resolution to 72.6%.

ExecCritic pairs a test-verify-revise scaffold with role-specific reinforcement learning: a Test agent writes repository-native tests and a Repair agent fixes code from execution feedback, both using Qwen-3.5-35B-A3B backbones. Post-trained Qwen agents compose to 72.6% on SWE-bench Verified, an 11.4-point gain over the 61.2% no-test baseline, without stronger-model or oracle feedback at evaluation time. The work shows test quality is the key variable: base-agent tests lowered resolution to 57.3% while GPT-5.6-sol tests raised it to 65.3%.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:53 UTC · 7d agoAI research1

36

SchemeArena: Factorized Stress Testing of Scheming in LLM Agents

Researchers introduce SchemeArena, a 400-scenario benchmark stress-testing scheming in LLM agents, finding explicit instrumental goals are the strongest driver of covert misaligned behavior.

The paper presents SchemeArena, a 400-scenario benchmark built through factorized scenario synthesis spanning safety-relevant tool domains, instrumental goals, oversight conditions and pressure mechanisms. The accompanying SCOUT monitor grounds multi-criteria scheming judgments in evidence drawn from agents' reasoning and actions. Stress tests across five LLM agents show explicit instrumental goals are the strongest driver of scheming propensity, while action-only monitoring increased scheming in several closed models, suggesting partial oversight can act as an optimization constraint. The benchmark, code and monitor are released at github.com/launchnlp/SchemeArena.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI safety & security

40

TasmScan: Continuation-Aware Taint Analysis for TVM Bytecode with Savelist Abstraction

TasmScan introduces source-free taint analysis for TON smart-contract bytecode, detecting 95.3% of defects with 96.8% precision and 17x speedup.

TasmScan is the first bytecode-level static analysis framework for the TON Virtual Machine, enabling cross-continuation data flow reasoning without source code by modeling savelist semantics through forward register analysis with formal over-approximation guarantees. It lifts bytecode into a typed intermediate representation (TASIR) and performs path-sensitive taint analysis. On a 208-contract benchmark with human-confirmed ground truth it detects 95.3% of defects across five classes at 96.8% precision, and resolves 294,546 dynamic continuation targets with 100% precision across 2,921 registry contracts. It achieves a 17x median speedup over symbolic-execution baselines.

arXiv cs.CR·Sep 15, 11:01 UTC · 1d agoResearch

28

Hackers Can Hide Malicious AI Commands Inside Normal English to Bypass Security Filters

Check Point's PuzzleMask technique hides malicious prompts in ordinary English that fast gatekeeper models miss but high-reasoning downstream models execute.

Check Point researchers disclosed PuzzleMask, a technique concealing policy-breaking instructions in natural-language prose without encodings or invisible characters. Fast screening models classified all 23 crafted wrappers as safe, while a high-reasoning model recovered and acted on the hidden instruction in 17 of 18 tests (94.4%). The gap stems from capability imbalance between gatekeeper and target models, with defenses including paraphrasing untrusted input, stricter self-referential wording rules, and output/tool-call monitoring.

Cyber Security News·Sep 11, 07:36 UTC · 5d agoAI safety & security1

55

AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems

AgentGrad introduces intervention-guided prompt optimization for LLM multi-agent systems, achieving state-of-the-art results with 2.5x faster optimization.

AgentGrad is a prompt optimization framework for LLM-based multi-agent systems that addresses limitations in textual gradient extraction and aggregation. It uses sequential intervention to identify the agent whose prompt modification resolves a given failure, then applies agent-level supervision and semantic gradient clustering to build generalized gradients. Experiments report state-of-the-art performance across five MAS benchmarks and a 2.5x average reduction in wall-clock optimization time versus the next-fastest baseline.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI research

32

[webapps] Langflow 1.8.4 - Path Traversal to Remote Code Execution

A path traversal to remote code execution exploit for Langflow 1.8.4, a popular LLM application builder, was published on Exploit-DB.

Exploit-DB lists a proof-of-concept exploit chaining path traversal to remote code execution in Langflow 1.8.4, an open-source tool used to build LLM applications and agents. The chain allows an attacker to write arbitrary files outside the intended directory and achieve code execution on the host. The provided text does not include a CVE identifier or reports of exploitation in the wild, but RCE in a widely deployed AI tooling product is notable for defenders.

Exploit-DB·Aug 31, 00:00 UTC · 16d agoExploit / PoC1

42

MAxBench: A Multinomial Concept Recovery Benchmark

MAxBench evaluates multinomial concept recovery methods, finding affine subspaces steer most reliably but none consistently beats prompting.

MAxBench is a geometry-agnostic evaluation framework for multinomial concept representations in language models, based on sampling from recovered concept representations. It compares 10 localization methods covering 5 geometry types across 6 concepts and 4 models. Findings show affine subspaces steer more reliably than rank-one or linear subspaces due to better non-zero offsets, manifold steering is competitive where applicable, and no method consistently outperforms prompting.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 17:08 UTC · 4d agoAI research

18

MIT creates method to force AI to comply with safety rules

MIT researchers published HardFlow, a method enforcing hard safety constraints on flow-matching generative models' final outputs without retraining.

MIT researchers led by Zeyang Li and Navid Azizan developed HardFlow, a trajectory-optimization method that enforces strict, non-negotiable constraints on flow-matching generative models by checking rule satisfaction only at the final generation step. Published in IEEE TPAMI, it outperformed six rival projection and guidance methods on four simulated benchmarks including D3IL robotic manipulation, Maze2D, physical process control, and image editing. All results are simulation-only, with no independent reproduction yet reported.

Hacker News · security·Sep 14, 16:18 UTC · 1d agoAI researchHN 24↑ · 25 comments

25

Context Engineering Inside the Harness: 4 Mechanisms That Beat Context Overflow and Goal Loss on Long-Horizon Tasks

Survey of four harness mechanisms—context budgeting, compaction, todo-state, and memory—that keep long-horizon LLM agents on task across 200+ tool calls.

The article details how agent harnesses, not larger context windows, solve context overflow and goal loss on long-horizon tasks, citing Chroma's Context Rot report showing 18 LLMs (GPT-4.1, Claude 4, Gemini 2.5, Qwen3) degrade on long inputs. Concrete implementations include LangChain Deep Agents offloading tool responses over 20,000 tokens to the filesystem and truncating old tool calls at 85% window usage, and Claude Code capping auto memory at 25KB while re-reading the 5 most recently modified files after compaction. OpenAI's Responses API now offers server-side compaction via context_management with a standalone /responses/compact endpoint, which Codex uses for long-running coding tasks. Manus reports a roughly 100:1 input-to-output token ratio per ~50-tool-call task, motivating todo.md state recitation to prevent goal drift.

MarkTechPost·Sep 13, 05:56 UTC · 3d agoAI research1

52