CALYREX: Cross-Attention LaYeR EXtended Transformers for System Prompt Anchoring

• URL: http://arxiv.org/abs/2605.09737v1
• Date: Sun, 10 May 2026 20:12:39 GMT
• Title: CALYREX: Cross-Attention LaYeR EXtended Transformers for System Prompt Anchoring
• Abstract summary: We propose CALYREX, which utilizes cross-attention between input and system prompt to structurally isolate and anchor the rule.<n>At 8B scale, controlling for training data, backbone, and parameter budget, CALYREX yields $+7.4%$ on instruction-following (IFEval) and $+16.3%$ on multi-turn instruction adherence.<n>This advantage appears to widen with model scale, consistent with larger models more effectively utilizing the dedicated routing pathway.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Modern large language models (LLMs) rely on system prompts to establish behavioral constraints and safety rules. Standard causal self-attention treats privileged instructions and untrusted user content with equal structural priority -- a mismatch that leaves models vulnerable to prompt injection and instruction erosion over extended contexts. We propose CALYREX (Cross-Attention LaYeR EXtended transformers), which utilizes cross-attention between input and system prompt to structurally isolate and anchor the rule. A placement ablation on a 1.5B backbone identifies insertion at the final eighth of layers as optimal, confirmed by mechanistic activation analysis showing behavioral constraints are naturally concentrated there. At 8B scale, controlling for training data, backbone, and parameter budget, CALYREX yields $+7.4\%$ on instruction-following (IFEval) and $+16.3\%$ on multi-turn instruction adherence, while reducing many-shot jailbreaking attack success rate by $13\%$. This advantage appears to widen with model scale, consistent with larger models more effectively utilizing the dedicated routing pathway.

Related papers

• CARE: Contrastive Anchor-based Rubric Evolution for Large Language Model Post-Training [19.798998515276523]
  Reinforcement learning decomposes open-ended instructions into prompt-specific, flexible rubrics.<n> static rubrics are inevitably hacked as the policy evolves.<n>We propose $textbfCARE$ which grounds every rubric evolution step in a high-quality anchor response.
  arXiv  Detail & Related papers  (2026-09-01T08:22:12Z)
• TACIT-Switch: Cost-Aware Model Escalation for LLM Agents from Censored Supervision [2.681967953218225]
  Agents with smaller language-model backbones can drift into persistent failure modes, whereas those with larger backbones are generally more reliable but more costly.<n>This reliability-cost trade-off motivates routing methods that decide when to invoke an agent with a larger backbone.<n>Our method, TACIT-SWITCH, learns permanent handoff policies from accumulated trajectory evidence and Teacher-Annotated Censored Intervention Times.
  arXiv  Detail & Related papers  (2026-08-28T04:33:11Z)
• Dynamic Coalition Formation and Communication Pricing in Skill-Based Agentic AI Systems [0.0]
  We model agent selection and communication as a cooperative game with task-conditioned net utility $U(Cmid x)=V(Cmid x)-sum_iin Cc_i$.<n>We propose a marginal-value activation rule and greedy router, extend the model to optimize communication edges with per-edge costs, and use estimated Shapley values to predict which agents are worth contacting before and during execution.
  arXiv  Detail & Related papers  (2026-07-24T19:25:21Z)
• Understanding-Enhanced Model Collaboration for Long-Tailed Egocentric Mistake Detection [85.49213290363834]
  We propose an Understanding-Enhanced Model Collaboration Method (UE-MCM)<n>It combines efficient coarse-grained video understanding with accurate fine-grained action reasoning.<n>The resulting system balances speed and accuracy, making it effective for detecting subtle, rare, and ambiguous mistakes in egocentric instructional videos.
  arXiv  Detail & Related papers  (2026-06-01T11:50:16Z)
• Jailbreaking the Matrix: Nullspace Steering for Controlled Model Subversion [12.201783188544093]
  Head-Masked Nullspace Steering (HMNS) identifies attention heads most causally responsible for a model's default behavior.<n>This is the first jailbreak method to leverage geometry-aware, interpretability-informed interventions.
  arXiv  Detail & Related papers  (2026-04-11T19:19:05Z)
• How Alignment Routes: Localizing, Scaling, and Controlling Policy Circuits in Language Models [0.0]
  This paper localizes the policy routing mechanism in alignment-trained language models.<n>An intermediate-layer attention gate reads detected content and triggers deeper heads that boost the signal toward refusal.
  arXiv  Detail & Related papers  (2026-04-06T03:20:37Z)
• Quantifying Self-Preservation Bias in Large Language Models [9.590157416396194]
  We introduce the emphTwo-role Benchmark for Self-Preservation.<n>It measures how often role identity overrides objective utility.<n>We observe that in low-improvement regimes, models exploit the interpretive slack to post-hoc rationalization.
  arXiv  Detail & Related papers  (2026-04-02T15:38:31Z)
• Breaking Contextual Inertia: Reinforcement Learning with Single-Turn Anchors for Stable Multi-Turn Interaction [49.03500737694832]
  We introduce textbfReinforcement textbfLearning with textbfTurn textbfRLSTA, a generalizable training approach designed to stabilize multi-turn interaction.<n>Experiments show that RLSTA significantly outperforms standard fine-tuning and abstention-based methods.
  arXiv  Detail & Related papers  (2026-03-05T04:04:59Z)
• Rethinking Multi-Condition DiTs: Eliminating Redundant Attention via Position-Alignment and Keyword-Scoping [61.459927600301654]
  Multi-condition control is bottlenecked by the conventional concatenate-and-attend'' strategy.<n>Our analysis reveals that much of this cross-modal interaction is spatially or semantically redundant.<n>We propose Position-aligned and Keyword-scoped Attention (PKA), a highly efficient framework designed to eliminate these redundancies.
  arXiv  Detail & Related papers  (2026-02-06T16:39:10Z)
• Phase Transition for Budgeted Multi-Agent Synergy [41.486076708302456]
  Multi-agent systems can improve reliability, yet under a fixed inference budget they often help, saturate, or even collapse.<n>We develop a minimal and calibratable theory that predicts these regimes from three binding constraints of modern agent stacks.
  arXiv  Detail & Related papers  (2026-01-24T05:32:50Z)
• ReasAlign: Reasoning Enhanced Safety Alignment against Prompt Injection Attack [52.17935054046577]
  We present ReasAlign, a model-level solution to improve safety alignment against indirect prompt injection attacks.<n>ReasAlign incorporates structured reasoning steps to analyze user queries, detect conflicting instructions, and preserve the continuity of the user's intended tasks.
  arXiv  Detail & Related papers  (2026-01-15T08:23:38Z)
• RISER: Orchestrating Latent Reasoning Skills for Adaptive Activation Steering [62.63376387138257]
  We propose a plug-and-play intervention framework that adaptively steers large language models (LLMs) reasoning in activation space.<n>RISER constructs a library of reusable reasoning vectors and employs a lightweight Router to dynamically compose them for each input.<n>The Router is optimized via reinforcement learning under task-level rewards, activating latent cognitive primitives in an emergent and compositional manner.
  arXiv  Detail & Related papers  (2026-01-14T08:04:33Z)
• Beyond 'Aha!': Toward Systematic Meta-Abilities Alignment in Large Reasoning Models [86.88657425848547]
  Large reasoning models (LRMs) already possess a latent capacity for long chain-of-thought reasoning.<n>We explicitly align models with three meta-abilities: deduction, induction, and abduction, using automatically generated, self-verifiable tasks.<n>Our three stage-pipeline individual alignment, parameter-space merging, and domain-specific reinforcement learning, boosts performance by over 10% relative to instruction-tuned baselines.
  arXiv  Detail & Related papers  (2025-05-15T17:58:33Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.