Skip to main content
Search

# Tool-Integrated Reasoning

4 related articles

Subscribe to updates
  1. arXiv

    Tool calls get worse with training: TACA supervises the reasoning-to-call transition and lifts six-benchmark accuracy by 4.61 points

    In multi-turn tool-use reinforcement learning, reasoning accuracy can rise early and then fall below the non-agentic baseline; TACA traces this to negative aggregate credit on tool-call tokens and instead supervises the step where reasoning turns into a tool call. Against MT-GRPO it raises mean accuracy over six reasoning benchmarks by 4.61 and 3.61 percentage points for Qwen2.5-1.5B and 3B, while tool execution success during training stays at 86.28%.

  2. Claude Platform opens OpenTelemetry export: trace Messages API calls and managed-agent sessions in your own backend

    Claude Platform now exports tracing data for Messages API requests and Claude Managed Agents sessions as OpenTelemetry spans, so developers can view those calls in their own observability backend.

  3. arXiv

    FlyBy lets a 4B small reasoning model query stronger models at knowledge bottlenecks, beating Qwen3-14B on 1,158 hard problems at 2.7x lower serving cost

    Through counterfactual interventions at intermediate reasoning states across two model families and multiple scales, the work finds that self-refinement largely consolidates probability mass onto solutions already reachable from the current state rather than making new ones reachable, distinguishes execution bottlenecks from knowledge bottlenecks, and introduces FlyBy, a selective querying framework that trains 4B and 8B models to reason first, diagnose what remains unresolved, and query stronger models at knowledge bottlenecks; on 1,158 hard problems across six benchmarks FlyBy-4B reaches 45.96% pass@8, surpassing Qwen3-14B's 41.64% at 2.7x lower serving cost, and FlyBy-8B raises pass@8 to 51.81%.

  4. PrefixAgent uses a two-phase LLM agent with e-graph trajectory fine-tuning to cut 64-bit prefix adder area to 938 µm², 11.3% below the strongest baseline

    The work proposes PrefixAgent, a two-phase large-language-model-driven framework for prefix adder optimization: in Phase I a fine-tuned large reasoning model iteratively optimizes the backbone via regroup tool calls, and in Phase II it performs local timing repair with level-opt, fanout-opt, and node clone tools; the authors use e-graph equality saturation and explanation to generate interpretable rewrite trajectories as supervision data, and under the NanGate45 and OpenROAD flow PrefixAgent produces smaller-area adders than DP, MCTS, PrefixRL, CircuitVAE, and PrefixGPT in nearly all configurations, achieving up to 11.3% area reduction over the best baseline at 64 bits and also improving area in a commercial flow and a 256-PE systolic array.

You’re all caught up