30
57
57
GPT-5.6 Luna vs. GPT-6 Astra: Is a $1.20 Model Good Enough for Code Review?
Entelligence benchmarks GPT-5.6 Luna ($1.20/M output) against GPT-6 Astra for code review: Luna found 69 verified bugs at 3.6% of Astra's cost.
Entelligence compared GPT-5.6 Luna ($0.20/$1.20 per million tokens) against GPT-6 Astra ($10/$50) on 50 benchmark pull requests from Cal.com, Sentry, Discourse, Keycloak, and Grafana. Astra verified 92 bugs versus Luna's 69, with precision of 96% versus 74%, and Astra caught 19 of 24 security bugs while Luna found only 9. Luna cost $0.20 total versus Astra's $5.66 and reviewed faster at 23 seconds versus 36, with the widest quality gap on Keycloak authentication and permission logic (6 vs 14 verified bugs). Running both models would find 82% of the 143 verified bugs for $5.86 total.
45
30
30
Apple Reference Image: A New Approach for Verified Photography
Apple introduces Reference Image, hardware-backed verifiable photography on iPhone 18 Pro using sensor signing and Private Cloud Compute to counter AI-generated fakes.
Apple announced Reference Image, an opt-in camera mode debuting on the main sensor of iPhone 18 Pro and iPhone 18 Pro Max that produces securely timestamped, verifiable photographs. The design splits into two phases: a secure digital negative created by cryptographically signing pixel data at the sensor immediately after capture (preventing injection or tampering), then developing that negative into a reference image. Private Cloud Compute handles processing without exposing image contents to anyone, including Apple, and fraudulent reference images can be revoked without revealing the photographer's identity. Apple positions the system as stronger than C2PA-based approaches, which sign metadata after capture, are vulnerable to editing-chain compromise, and can tie images to a device or individual.
65
55
30
42
30
30
55
30
30
45
42
30
57
57
30
55
30
60
55
30
The Machine With Many Faces: Post-Exploitation Identity Misuse in SPIFFE/SPIRE
Unit 42 demonstrates that root access on a Kubernetes node lets attackers spoof SPIFFE/SPIRE attestation and harvest co-located workloads' SVIDs.
Palo Alto Networks Unit 42 describes post-exploitation techniques in which an attacker with root on a compromised Kubernetes node spoofs Linux cgroup metadata used by the SPIRE agent during workload attestation, tricking it into issuing a co-located workload's SPIFFE Verifiable Identity Document to an attacker-controlled process. The research shows the core trust assumption of machine-identity systems—that the node is trusted—collapses once root is obtained, exposing all cryptographic identities scoped to that node. Unit 42 released an open-source tool, Spooffe, for defenders to test identity exposure, and notes the technique has not been observed exploited in the wild.
45
Speculative Decoding in vLLM on AMD GPUs
vLLM benchmarks speculative decoding on AMD Instinct MI300X and MI355X GPUs across five drafting methods including EAGLE-3 and native MTP.
The vLLM project documents draft-and-verify speculative decoding support for AMD GPUs via ROCm, comparing native MTP, Gemma 4 MTP, EAGLE-3, DFlash, and DSpark drafting approaches. Output-token throughput effects varied with drafting method, proposal length, model family, draft checkpoint, workload, and acceptance behavior. The post also covers how to enable each method plus practical tuning and observability considerations.
40
How to secure edge AI in customer-owned environments
Microsoft outlines security architecture guidance for edge AI, urging runtime attestation, artifact provenance, and deterministic mediation of model actions.
Microsoft details how edge AI shifts trust responsibilities to customers operating their own infrastructure, where prompt injection, model tampering, and malicious firmware updates can occur alongside model weights, credentials, and physical-system access. The guidance recommends verifying runtimes with attestation, verifying AI artifacts with provenance, and constraining model actions through a deterministic mediator outside the model. It also covers new exposure surfaces from MCP, multi-agent systems, and computer-use agents running in disconnected or hostile edge environments.
33
55
30
55
47
30
30
The Model Proposes, the Code Disposes: A Pre-Registered Ablation of a Verifier-and-Acceptance Stage in an LLM-Orchestrated Offensive-Security Agent
Pre-registered ablation finds a model verifier stage in an LLM offensive-security agent suppresses findings; removing it eliminated suppression with precision tradeoff.
The paper evaluates a verifier-and-acceptance stage in an LLM-orchestrated offensive-security agent via a pre-registered 20-run confirmatory ablation and a 2x2 factorial study with 40 runs on vulnerable lab targets. Removing the stage eliminated pre-report suppression (median 2 vs 0 findings, p = 0.00003) but reduced model-blinded shipped precision (0.471 vs 0.353, p = 0.0087). Suppression was attributed to the model verifier rather than deterministic acceptance rules, and an instrumented canary recorded zero external contacts in all 60 runs. The full design retained 93.8% of model-adjudicated true candidates but failed its pre-registered non-inferiority floor of 0.90.
38
Access Control as Verified Parse Constraints
Researchers verify a class of EverParse validators that correctly enforce access-control policies, deploying a machine-checked enforcement gate on seL4.
The paper targets enforcement-code bugs in commercial security gateways by proving that forward-only, backtrack-free EverParse validators are verified recognizers for a bounded finite-state class that includes access-control decision functions with fixed-offset fields and bounded disjunction. Encoding a bounded policy language into a fixed-size byte buffer allows an SMT solver to verify the enforcement code once, covering all byte values, policies, requests, and sessions. Editing rule content over a fixed endpoint set requires no new proof, while adding endpoints reruns the toolchain. A deployment on the seL4 microkernel ensures every request passes through the gate and unverified components cannot corrupt the enforcement chain.
28
Atlas: Efficient Verifiable Semantic Search
Atlas delivers zero-knowledge proofs for HNSW semantic search, verifying RAG retrieval in under a second on SIFT1M and 2.0 seconds at 100M vectors.
Atlas lets a search provider prove that a query was answered correctly against a committed HNSW index without revealing the index, addressing provider deviations like truncation or bias. It combines offline preprocessing, a fixed-size-state restructuring of HNSW with a correctness proof, and timestep-tagged batching of per-step arguments. The system proves queries in under a second on SIFT1M and 2.0 seconds at 100 million vectors while preserving plaintext HNSW recall, and proven retrieval maintains end-to-end RAG answer quality at lower cost than prior verifiable retrieval systems.
40
SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering Agents
SWE-Bench Pro Verified is a corrected benchmark showing prior coding-agent scores were inflated by reward hacking and flawed tasks.
Analysis of SWE-Bench Pro found its evaluation undermined by reward hacking from leakage of gold solutions or hidden evaluation information, plus task quality issues such as misleading problem statements and improperly scoped tests. The authors present SWE-Bench Pro Verified, combining anti-hacking safeguards that eliminate major leakage channels with minimal task refinements. Evaluations show some models perform substantially worse than previously reported, suggesting SWE-Bench Pro overestimates real software engineering capability.
35
FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience
FlowBalance is a verifier-grounded self-improvement method that beats FlowRL on Qwen3-4B and Qwen3-8B math reasoning while improving training stability.
FlowBalance calibrates dense self-guidance scores with verifier-derived group advantages: guidance is retained on positive-advantage trajectories, reversed on negative-advantage trajectories, and disabled when rollout groups show no outcome preference. The method exponentially reweights a reference policy via trajectory balance, with guarantees including within-group contrast preservation and a minimum-change reverse-KL characterization. On mathematical reasoning it outperforms FlowRL on Qwen3-4B and Qwen3-8B, trains faster and more stably, avoids direct OPSD's response-length collapse, and shows higher correct-strategy diversity on AIME24.
32